Elevated p53 Activities Restrict Differentiation Potential of MicroRNA-Deficient Pluripotent Stem Cells.

Pluripotent stem cells (PSCs) deficient for microRNAs (miRNAs), such as Dgcr8-/- or Dicer-/- embryonic stem cells (ESCs), contain no mature miRNA and cannot differentiate into somatic cells. How miRNA deficiency causes differentiation defects remains poorly understood. Here, we report that miR-302 is sufficient to enable neural differentiation of differentiation-incompetent Dgcr8-/- ESCs. Our data showed that miR-302 directly suppresses the tumor suppressor p53, which is modestly upregulated in Dgcr8-/- ESCs and serves as a barrier restricting neural differentiation. We demonstrated that direct inactivation of p53 by SV40 large T antigen, a short hairpin RNA against Trp53, or genetic ablation of Trp53 in Dgcr8-/- PSCs enables neural differentiation, while activation of p53 by the MDM2 inhibitor nutlin-3a in wild-type ESCs inhibits neural differentiation. Together, we demonstrate that a major function of miRNAs in neural differentiation is suppression of p53 and that modest activation of p53 blocks neural differentiation of miRNA-deficient PSCs.

Introduction

MicroRNAs (miRNAs) are small, non-coding RNAs that play important roles in embryogenesis, lineage specification, tissue homeostasis, and human diseases such as neurodegeneration and cancers (Abe and Bonini, 2013, Ivey and Srivastava, 2010, Lee and Dutta, 2009). Biogenesis of miRNAs requires the DROSHA-DGCR8 microprocessor, which cleaves pri-miRNAs into pre-miRNAs, and DICER, which further processes pre-miRNAs into mature miRNAs. Mature miRNAs are then incorporated into the RNA-induced silencing complex (RISC) to destabilize and/or suppress translation of target mRNAs (Ha and Kim, 2014). Notably, Dgcr8 or Dicer embryonic stem cells (ESCs), which have complete miRNA loss, are unable to differentiate into mature somatic cells, indicating that miRNAs are required for differentiation (Kanellopoulou et al., 2005, Liu et al., 2015, Murchison et al., 2005, Wang et al., 2007). Although a number of miRNAs have been identified to facilitate differentiation of wild-type pluripotent stem cells (PSCs) by suppressing pluripotency-promoting genes and/or instructing lineage specification by forming feedforward and feedback regulatory loops with transcriptional regulators (Ivey and Srivastava, 2010, Wang et al., 2007, Xu et al., 2009), no miRNA has yet been reported to rescue the differentiation defects of miRNA-deficient PSCs. Therefore, it remains poorly understood how the lack of miRNAs eliminates the differentiation capacity of PSCs, and which miRNAs are essential to confer differentiation competence.

In this study, we investigated this question by testing the hypothesis that certain miRNAs, most likely those abundantly expressed in PSCs or immediate progenitors, confer differentiation competence to PSCs. We first developed a strategy that allows stable expression of individual miRNAs in miRNA-deficient Dgcr8−/− ESCs. Using this strategy, we tested the function of candidate miRNAs in an in vitro neural differentiation assay. We chose neural differentiation because our previous data demonstrated that, although incapable of producing any differentiated lineages, embryoid bodies (EBs) formed by Dgcr8−/− induced pluripotent stem cells (iPSCs) modestly express markers of neuroectoderm but not mesoderm or endoderm (Liu et al., 2015), which suggested that fewer barriers likely need to be overcome during the differentiation of Dgcr8−/− PSCs into neurons. Our data demonstrated that expression of miR-302, which is best known for promoting pluripotency (Melton et al., 2010, Subramanyam et al., 2011, Tiscornia and Izpisua Belmonte, 2010, Wang et al., 2008, Wang et al., 2013), enabled neural differentiation of Dgcr8−/− ESCs. We discovered that miR-302 directly suppresses the tumor suppressor p53, which is expressed at a higher basal level in Dgcr8−/− ESCs and blocks neural differentiation. Furthermore, direct inactivation of p53 by SV40 large T (LT) antigen, a short hairpin RNA (shRNA) against Trp53, or genetic ablation of Trp53 enabled neural differentiation of Dgcr8−/− PSCs, while activation of p53 in wild-type ESCs by the MDM2 inhibitor nutlin-3a inhibited neural differentiation. Together, our data demonstrate that p53 is a differentiation barrier of miRNA-deficient PSCs and a major function of miRNAs in neural lineage specification is suppression of p53.

Results

Expression of miR-302 Enables Neural Differentiation of Dgcr8 ESCs

Because Dicer or Dgcr8 PSCs can self-renew but cannot differentiate (Kanellopoulou et al., 2005, Liu et al., 2015, Murchison et al., 2005, Wang et al., 2007), we hypothesized that certain miRNAs, most likely those abundantly expressed in PSCs or immediate progenitors, confer differentiation competence to PSCs. To identify such miRNAs, we expressed mimics of candidate miRNAs into Dgcr8 ESCs and evaluated the differentiation potential of the resulting cells in an in vitro neural differentiation assay (Figures 1A–1C). The top candidate miRNAs included let-7, which induces pluripotency exit (Melton et al., 2010); miR-124 and miR-9, which promote neurogenesis (Kawahara et al., 2012); and miR-302, which is abundantly expressed in PSCs and early neural tissues (Parchem et al., 2014, Parchem et al., 2015).

Figure 1

Expression of miR-302 Mimics Enabled Neural Differentiation of Dgcr8 ESCs

(A–C) Immunostaining of neuron-specific markers TUJ1 (green) and MAP2 (red) in embryoid bodies (EBs) formed by (A–A″) wild-type, (B–B″) Dgcr8-shctrl, and (C–C″) Dgcr8-302 ESCs. Scale bars, 50 μm.

(D) Expression strategy for sh-miR-302 in Dgcr8−/− cells. (D) Sequence alignment of endogenous miR-302 and sh-miR-302. Sequences that are aligned are color labeled (green) and seed sequences are shaded. (D′) Hairpin structure of sh-miR-302. Red arrows represent the predicted Dicer cutting sites. Percentage of correctly processed sh-miRs, which mimics the first 17 nt of the 5′ end of the endogenous miR-302, is presented (see also Table S1).

(E) Percentage (%) of TUJ1+ neuronal cells in total cells (DAPI staining, not shown) of EBs formed in (A–C). Error bars, SD; n = 3 independent biological repeats.

(F) qPCR analyses of known miR-302 targets Tgfbr2 and Lats2 (normalized to β-actin) in wild-type, Dgcr8-shctrl, and Dgcr8-302 ESCs. Error bar, SD; n = 3 independent biological repeats.

(G) Cell-cycle profile analysis of Dgcr8-shctrl and Dgcr8-302 ESCs by DNA content measurement (propidium iodide staining); n = 4, independent biological repeats.

(H–L) Immunostaining of neuron-specific markers TUJ1 (green) and (H) NeuN (red) and DAPI (blue), (I) Synapsin (red), (J) tyrosine hydroxylase (TH) (red), (K) vesicular glutamate transporter 1 (VGLU1) (red), and (L) γ-aminobutyric acid (GABA) (red) in EBs formed by Dgcr8-302 ESCs. Scale bars, 50 μm (white), 20 μm (blue).

(M and N) H&E staining of teratomas formed by (M) Dgcr8 ESCs, which contained predominantly undifferentiated cells, and (N) Dgcr8-302 ESCs, which contained many neuroepithelia. Scale bars, 50 μm. n = 5 independent biological repeats (see also Figure S2).

Several studies have identified miRNAs that regulate proliferation, self-renewal, pluripotency exit, and DNA methylation of PSCs by transiently transfecting chemically synthesized miRNA mimics into Dgcr8−/− or Dicer−/– ESCs (Benetti et al., 2008, Melton et al., 2010, Sinkkonen et al., 2008, Wang et al., 2008). However, because lineage specification often involves multiple rounds of cell division, or takes place over days or even weeks, a strategy that allows stable expression of miRNA mimics would be advantageous. We developed such as method by inserting mature miRNA sequences into lentivirally delivered shRNAs (sh-miRs), which structurally resemble pre-miRNAs and therefore bypass DGCR8 for miRNA biogenesis (Figures 1D and S1). Transcription of shRNAs relies on RNA Pol III-dependent promoters (e.g., the U6 promoter), which preferentially initiate transcription from a guanine (G) residue (Goomer and Kunkel, 1992, Kunkel et al., 1986). However, most RISC-associated mature miRNAs do not have a G residue at the 5′ end (Hu et al., 2009), which makes the 3′ arm of shRNA more suitable than the 5′ arm for miRNA expression. Because the seed sequence, which is between position 2 and 7 on mature miRNAs, is the major determinant for mRNA target selection (Ha and Kim, 2014), the exact sequence and target specificity of a 3′ arm miRNA is determined by the cleaving site of DICER (Figure 1D'). Based on knowledge of how DICER processes stem-loop structures (Gu et al., 2012), we inserted mature miRNA sequences into the 3′ arm starting from the third nucleotide position and kept the stem of the hairpin 21 nucleotides in length (Figure 1D'). Such a design ensures proper processing of the inserted miRNA by DICER.

To validate the neural differentiation assay, we first differentiated wild-type and Dgcr8 ESCs expressing a control shRNA (Dgcr8-shctrl). After 14 days of differentiation in EBs, TUJ1+ and MAP2+ neurons were evident in wild-type (Figure 1A–A″) but not Dgcr8-shctrl ESCs (Figure 1B–B″), confirming that Dgcr8 PSCs were defective in differentiation (Liu et al., 2015, Wang et al., 2007). Among the tested miRNA mimics (Figures 1 and S2), we discovered that Dgcr8 ESCs expressing sh-miR-302 (Dgcr8-302) can efficiently differentiate into TUJ1+ and MAP2+ neurons (Figures 1C–C″ and 1E). RNA sequencing (RNA-seq) analysis demonstrated that 98.9% of miRNAs processed from sh-miR-302 mimicked the endogenous miR-302 (Figure 1D; Table S1), which supports the previous report that small RNAs can be precisely expressed and processed by DICER when inserted into well-designed hairpins (Gu et al., 2012). qPCR analysis confirmed that the known miR-302 mRNA targets Tgfbr2 and Lats2, which were de-repressed in Dgcr8-shctrl ESCs, were repressed in Dgcr8-302 ESCs (Figure 1F). Cell-cycle progression from G1 to S phase was significantly accelerated in Dgcr8-302 ESCs (Figure 1G), consistent with the report that miR-302 accelerate the G1-S phase transition (Wang et al., 2008).

Neurons differentiated from Dgcr8-302 ESCs expressed the mature neuronal markers NeuN, a post-mitotic neuron-expressed nuclear antigen (Figure 1H), and Synapsin, a membrane protein on synaptic vesicles (Figure 1I). Neurons positive for tyrosine hydroxylase, vesicular glutamate transporter 1, and γ-aminobutyric acid (GABA) (Figures 1J–1L), which are markers for dopaminergic, glutamatergic, and GABAergic neurons, respectively, can also be detected with extended Dgcr8-302 EB differentiation (28 days). When injected subcutaneously into immunodeficient mice, teratomas formed by Dgcr8 ESCs contained predominantly undifferentiated cells (Figure 1M), as reported previously (Liu et al., 2015, Wang et al., 2007), whereas teratomas formed by Dgcr8-302 ESCs consisted of many neuroepithelial tissues (Figure 1N). However, we did not detect obvious mesodermal and endodermal tissues in teratomas formed by Dgcr8-302 ESCs, suggesting that differentiation into these lineages requires miRNAs other than miR-302.

We attributed the observed rescue of neural differentiation in Dgcr8 ESCs to functions specific to miR-302. Indeed, expression of let-7, which induces pluripotency exit of Dgcr8 ESCs (Melton et al., 2010), or of miR-9 and miR-124, two known neurogenesis-promoting miRNAs (Kawahara et al., 2012), failed to rescue the differentiation defect (Figures S2A–S2C). Confirming that the expressed miRNAs were functional, expression of let-7b led to pluripotency exit of Dgcr8 ESCs as reported by Melton et al. (2010) (Figure S2D–S2D″), while miR-9 and miR-124 downregulated expression of known mRNA target genes (Figures S2E and S2F).

Inhibition of TGF-β and BMP Pathways in Dgcr8−/− ESCs Cannot Rescue the Neural Differentiation Defect

Identification of the mRNA targets of miR-302 is critical to understanding how miRNAs regulate the differentiation of PSCs. Among the known miR-302 targets are Tgfbr2 (Figure 1F), a receptor mediating transforming growth factor-β (TGF-β) signaling, and genes within the bone morphogenetic protein (BMP) signaling pathway (Lipchina et al., 2011, Subramanyam et al., 2011). Because inhibition of TGF-β and BMP pathways induces efficient neural differentiation (Chambers et al., 2009), we tested whether sh-miR-302 enabled neural differentiation of Dgcr8 ESCs by repressing these pathways. We demonstrated that inhibition of the TGF-β pathway with the chemical inhibitor SB431542 and/or inhibition of the BMP pathway by Noggin in Dgcr8 ESCs had little effect on neural differentiation (Figures 2A–2D), and therefore could not fully account for the effect of sh-miR-302 expression (Figures 2E and 2F).

Figure 2

Inhibition of BMP and TGF-β Signaling in Dgcr8 ESCs Cannot Rescue the Neural Differentiation Defect

(A–E) Immunostaining of neuron-specific markers TUJ1 (green) and MAP2 (red) in EBs formed by (A–D) Dgcr8-shctrl or (E) Dgcr8-302 ESCs. During differentiation, EBs of Dgcr8-shctrl ESCs were treated with (B) Noggin (250 ng/mL), a BMP signaling inhibitor, (C) SB431542 (15 μM), a TGF-β signaling inhibitor, or (D) both.

(F) Percentage (%) of TUJ1+ neuronal cells in total cells (DAPI staining, not shown) of EBs formed in (A–E). Note that inhibition of BMP and/or TGF-β signaling had little effect on induction of neural differentiation, and could not account for the observed sh-miR-302-mediated rescue. Error bars, SD; n = 3 independent biological repeats.

p53 Is a Differentiation Barrier in PSCs

To gain insight into how miR-302 enables neural differentiation of Dgcr8 ESCs, we compared expression profiles of Dgcr8-shctrl and Dgcr8-302 ESCs by microarray analysis. Expression of sh-miR-302 resulted in upregulation of 643 genes and downregulation 941 genes by 1.5-fold (p < 0.05) in Dgcr8 ESCs (Figures 3A and 3B; Table S2). Gene set enrichment analysis (GSEA) revealed downregulation of multiple gene sets in Dgcr8−/−-302 ESCs (Table S3), including genes regulated by p53 or induced by conditions related to p53 activation, such as apoptosis (Figures 3C and 3D). These data led us to hypothesize that increased p53 activity, which is normally suppressed by miR-302, blocks neural differentiation in Dgcr8−/− ESCs.

Figure 3

Expression Profiling Reveals that sh-miR-302 Suppresses p53 Target Genes in Dgcr8 ESCs

(A) Unsupervised clustering analysis segregates biological repeats of Dgcr8-shctrl ESCs (n = 3) from Dgcr8-302 ESCs (n = 3 independent biological repeats).

(B) Scatterplot showing that ectopic expression of miR-302 leads to upregulation of 643 genes and downregulation of 942 genes by 1.5-fold (p < 0.05) in Dgcr8 ESCs. Green dots represent the significantly differentially expressed genes between the Dgcr8−/−-shctrl and Dgcr8-302 ESCs groups (see also Table S2).

(C) Gene set enrichment analysis (GSEA) reveals that genes regulated by p53 or upregulated in apoptosis are repressed in Dgcr8 ESCs by sh-miR-302 (see also Table S3).

(D) Heatmap showing differential expression of selected genes between Dgcr8-shctrl and Dgcr8-302 ESCs.

To test this hypothesis, we ectopically expressed the SV40 LT antigen, which inactivates both p53 and all RB family proteins (An et al., 2012) in Dgcr8 ESCs (Dgcr8-LT). TUJ1+, MAP2+, and NeuN + neurons were readily detected in EBs formed by Dgcr8-LT ESCs (Figures 4A and 4B). To separate the roles of p53 and the RB family proteins, we ectopically expressed an shRNA against p53 (Dgcr8-shp53) or T121 (Dgcr8-T121), a truncated version of the LT antigen that only inactivates RB proteins (Ewen et al., 1989), in Dgcr8 ESCs. Neurons expressing TUJ1, MAP2, and NeuN were evident in EBs formed by Dgcr8-shp53 ESCs (Figures 4C and 4D), but were rarely seen in EBs formed by Dgcr8-T121 ESCs (Figure 4E). qPCR analysis verified a reduction of Trp53 mRNA in Dgcr8-shp53 ESCs to 25%–30% of that in Dgcr8-shctrl ESCs (Figure 4F), suggesting that an increase in p53 by 3- to 4-fold could block differentiation of Dgcr8 ESCs. In addition, teratomas of Dgcr8−/−-shp53 ESCs contained primarily neural tissues but not mesodermal and endodermal derivatives (Figure 4G).

Figure 4

Elevated p53 Activity Serves as a Barrier Blocking Differentiation of Dgcr8 and Wild-Type PSCs

(A–E) Immunostaining of neuron-specific markers TUJ1 (green), MAP2 (red in A, C, and E), and NeuN (red in B and D) in EBs formed by Dgcr8 ESCs expressing (A and B) the SV40 large T antigen (Dgcr8-LT), (C and D) an shRNA against p53 (Dgcr8-shp53), or (E) T121 (Dgcr8-T121), a truncated large T antigen that only inactivates RB family proteins. Scale bars, 50 μm.

(F) qPCR analysis of p53 in Dgcr8-shctrl and Dgcr8-shp53 ESCs confirms p53 knockdown. Error bars, SD; n = 3 independent biological repeats.

(G) H&E staining of teratomas formed by Dgcr8-shp53 ESCs, which contain many neuroepithelia. Scale bars, 50 μm. n = 3 independent biological repeats.

(H and I) Immunostaining of neuron-specific markers TUJ1 (green) and (H) MAP2 (red) and (I) NeuN (red) in EBs formed by Dgcr8−/−; Trp53−/− iPSCs. Scale bars, 50 μm.

(J) Immunoblotting of p53 and GAPDH in wild-type ESCs treated with DMSO or nutlin-3a.

(K and L) Immunostaining of TUJ1 (green) and MAP2 (red) in EBs formed by wild-type ESCs treated with (K) DMSO or (L) nutlin-3a. Scale bars, 50 μm.

We also generated iPSCs deficient for both Dgcr8 and Trp53 (Dgcr8−/−;Trp53−/− iPSCs) from tail tip fibroblasts of Dgcr8flox/flox;Trp53flox/flox mice. Consistently, Dgcr8−/−;Trp53−/− iPSCs efficiently differentiated into TUJ1+, MAP2+, and NeuN + neurons (Figures 4H and 4I), in contrast to Dgcr8−/− iPSCs, which were differentiation incompetent (Liu et al., 2015). Furthermore, wild-type ESCs treated with nutlin-3a (Figure 4J), a small molecule that stabilizes p53 (Vassilev et al., 2004), exhibited greatly reduced neural differentiation (Figures 4K and 4L). Together, these data demonstrate that elevated p53 activities block neural differentiation of PSCs.

p53 Is Directly Suppressed by miR-302

To explore the mechanism by which p53 is regulated by miR-302, we conducted a computational search and identified a putative miR-302 recognition site on the 3′ UTR of p53 (Figure 5A). Although mRNA of Trp53 is expressed at similar levels (Figure 5B), we found that Dgcr8−/− ESCs express approximately 3- to 4-fold higher levels of p53 protein than the wild-type or Dgcr8−/−-302 ESCs (Figure 5C). To test the functionality of the miR-302 recognition site, we performed a luciferase reporter assay. We generated luciferase reporters containing the intact 3′ UTR of p53 (WT-Luc) or a mutant 3′ UTR (Mut-Luc), which contains two mutated nucleotides within the seed sequence of the miR-302 recognition site (Figure 5D). We also generated a positive control reporter (302-Luc) by inserting a sequence fully complementary to miR-302d. miR-302 effectively suppressed luciferase activities of the 302-Luc and WT-Luc reporters, but not the Mut-Luc reporter (Figure 5E), which indicates that the miR-302 recognition site is functional and p53 is directly suppressed by miR-302.

Figure 5

miR-302 Directly Suppresses p53 Expression

(A) Computational analysis predicting a miR-302 recognition site on the 3′ UTR of p53. The nucleotides mutated in the Mut-Luc reporter are underlined.

(B) qPCR analysis demonstrating that Trp53 is expressed at similar levels in wild-type, Dgcr8-shctrl, and Dgcr8-302 ESCs. Error bars, SD; n = 3–6 independent biological repeats.

(C) Immunoblotting of p53 and GAPDH in wild-type, Dgcr8, and Dgcr8-302 ESCs. Quantification of p53 was determined by ImageJ (normalized to GAPDH). n = 3 independent biological repeats.

(D) Sequences of the miR-302 recognition site on luciferase reporters. WT-Luc and Mut-Luc, the intact and mutant 3′ UTR of p53, respectively. The mutated nucleotides are shown in red.

(E) Luciferase reporter assay. Vector, empty pSicheck2 luciferase vector; 302-Luc, positive control reporter with sequence fully complementary to miR-302d. n = 3 independent biological repeats; ∗p < 0.05; ∗∗p < 0.01; Student's t test.

p53 Is Upregulated in Dgcr8−/− ESCs

Because p53 activation often leads to apoptosis, and PSCs are known to be sensitive to genotoxic insult (Aladjem et al., 1998, Hong and Stambrook, 2004, Roos et al., 2007), it was puzzling that Dgcr8 ESCs could tolerate elevated p53 activity. We found that p53 can be further induced in Dgcr8 ESCs by the DNA-damaging reagent neocarzinostatin (NCS) (Figure 6A). While wild-type ESCs did not undergo an obvious cell-cycle arrest upon NCS treatment, which agrees with previous reports (Aladjem et al., 1998, Qin et al., 2007), NCS-treated Dgcr8 ESCs exhibited clear cell-cycle arrest, as demonstrated by the marked reduction of bromodeoxyuridine (BrdU)-labeled S phase cells and accumulation of G2/M phase cells (Figures 6B and 6C). Interestingly, NCS treatment had little effect on the cell cycle of Dgcr8−/−-302 ESCs, which is similar to what was seen in wild-type ESCs (Figures 6B and 6C). Compared with wild-type and Dgcr8−/−-302 ESCs, induction of p53 by NCS is more efficient in Dgcr8−/− ESCs (Figure 6A), likely due to the lack of miR-302-mediated p53 suppression. Because the biological outcomes of p53 activation may be determined by the magnitude of activated p53 (Purvis et al., 2012), it is likely that the higher p53 levels in Dgcr8−/− ESCs contribute to the differences in cell-cycle regulation.

Figure 6

p53 Is Moderately Activated in Dgcr8 ESCs

(A) Immunoblotting of p53 in wild-type, Dgcr8−/−, and Dgcr8−/−-302 ESCs treated with neocarzinostatin (NCS) for the indicated times. Immunoblotting of GAPDH served as a loading control.

(B) NCS treatment causes cell-cycle arrest in Dgcr8−/− ESCs. Shown are representative fluorescence-activated cell sorting (FACS) plots of BrdU pulse-labeled ESCs 24 hr after DMSO (control) or NCS treatment in wild-type, Dgcr8−/−, and Dgcr8−/−-302 ESCs.

(C) Quantification of NCS-induced cell-cycle arrest. Error bars, SD; n = 4 independent biological repeats.

Next, we investigated how p53 level regulates the dual roles of p53 in ESCs: differentiation suppression and apoptosis induction. We first induced p53 expression in wild-type ESCs in a nutlin-3a dose-dependent manner (Figure 7A), and examined apoptosis of the nutlin-3a-treated ESCs. We found that increased levels of p53 had little effect on apoptosis rates of the nutlin-3a-treated ESCs, as determined by Annexin V staining (Figures 7B and 7C). This was surprising because ESCs are known to be sensitive to genotoxic insults (Aladjem et al., 1998, Hong and Stambrook, 2004, Roos et al., 2007). We then examined how ESCs respond to the DNA-damaging chemical NCS. We found that NCS strongly induced apoptosis in ESCs (Figures 7B and 7C), although NCS-induced p53 expression did not reach the levels induced by high doses of nutlin-3a (Figure 7A). These data demonstrated that apoptosis does not correlate well with p53 level in ESCs. In agreement with our data, Aladjem et al. (1998) reported that p53-null ESCs undergo apoptosis at a similar rate as wild-type ESCs in response to DNA damage, demonstrating that DNA damage-induced apoptosis may be executed via a p53-independent mechanism in ESCs. Because nutlin-3a is known to induce p53 by blocking MDM2-mediated p53 degradation without inducing DNA damage (Vassilev et al., 2004), our data suggested that increased p53 expression without DNA damage does not efficiently induce apoptosis in ESCs (Figures 7A–7C); however, such an increase strongly inhibited differentiation (Figures 4J–4L). Taken together, our data demonstrated that activation of p53, which is normally suppressed by miRNAs such as miR-302, serves as a barrier that restricts neural differentiation of PSCs (Figure 7D).

Figure 7

Increased p53 Expression without DNA Damage Does Not Efficiently Induce Apoptosis in ESCs

(A) Immunoblotting of p53 in ESCs treated with NCS or increasing concentrations of nutlin-3a. GAPDH was used as a loading control.

(B) Apoptosis of wild-type ESCs treated with NCS or nutlin-3a. Shown are representative FACS plots of Annexin V staining. DMSO-treated ESCs were used as a negative control.

(C) Quantification of apoptosis of wild-type ESCs treated with NCS or nutlin-3a. Error bars, SD; n = 3–7 independent biological repeats. ∗∗p < 0.01; N.S., not significant; two-tailed Student's t test.

(D) A model illustrating that miR-302 directly suppresses p53 expression and elevated p53 blocks neural differentiation.

Discussion

The miR-302 family, which is most abundantly expressed in epiblast-derived pluripotent stem cells and human ESCs (Card et al., 2008, Parchem et al., 2014, Suh et al., 2004), is generally believed to promote pluripotency by facilitating rapid cell-cycle progression and antagonizing the differentiation-inducing activities of let-7 miRNAs (Melton et al., 2010, Wang et al., 2008). Furthermore, miR-302 enhances iPSC derivation when co-expressed with the Yamanaka factors (Leonardo et al., 2012). In this study, we demonstrated that miR-302 promotes lineage specification by inhibiting p53 activation. Interestingly, inhibition of p53 has also been shown to enhance iPSC derivation (Spike and Wahl, 2011). Our data suggest that, among the many mechanisms that promote reprogramming, miR-302 may also facilitate reprogramming by suppressing p53 expression. Several other pluripotency transcriptional regulators, including OCT4, SOX2, and NANOG, have also been demonstrated to promote lineage specification (Frum et al., 2013, Graham et al., 2003, Messerschmidt and Kemler, 2010). These data therefore suggest that it is a general attribute of pluripotency regulators to play dual roles (i.e., promoting self-renewal and inducing lineage specification) in different cellular contexts.

Dicer or Dgcr8 mice embryos die before formation of body axis, underscoring an essential role of miRNAs in early embryogenesis (Bernstein et al., 2003, Wang et al., 2007). Tissue-specific disruption of Dicer or Dgcr8 has been performed to evaluate the roles of miRNAs in a variety of tissues, such as cerebral cortex and neuron types (Choi et al., 2008, Davis et al., 2008, De Pietri Tonelli et al., 2008, Kawase-Koga et al., 2009, Kim et al., 2007, Makeyev et al., 2007), neural crest (Chapnik et al., 2012, Nie et al., 2011), hair follicles (Andl et al., 2006), limb mesoderm (Harfe et al., 2005), and immune cells (Cobb et al., 2005). Interestingly, phenotypes associated with p53 activation, such as reduced proliferation and/or increased apoptosis, have been observed in virtually all such studies. Whether p53 activation contributes to the observed phenotypic defects in all these tissues remains an open question. Furthermore, we observed that teratomas formed by Dgcr8−/−-302 or Dgcr8−/−-shp53 ESCs contained primarily neural tissues but no obvious mesodermal and endodermal lineages. These data suggest that a major role of miRNAs in PSC differentiation into the neural lineage is to suppress p53 activation, while additional p53-independent functions of miRNAs are likely necessary for the differentiation of mesodermal and endodermal tissues.

Several lines of evidence suggest that, on the molecular level, p53 activation suppresses self-renewal and induces exit from pluripotency. p53 directly inhibits Nanog expression, suggesting a mechanism to eliminate damaged cells from the stem cell pool by inducing differentiation (Lin et al., 2005). Furthermore, p53 activated by DNA damage was demonstrated to bind promoters and enhancers of hundreds of genes, with a tendency of repressing genes for self-renewal and activating genes for differentiation (Li et al., 2012). However, whether these p53-induced transcriptional changes are sufficient to promote differentiation of PSCs has not been demonstrated. In this study, we found that miRNA deficiency moderately activates p53, which is sufficient to block differentiation of PSCs. Because the biological outcomes of p53 activation may be determined by the types of stress signals and the magnitude and duration of activated p53 (Horn and Vousden, 2007, Kastenhuber and Lowe, 2017, Khoo et al., 2014, Purvis et al., 2012), it is also likely that the moderately upregulated basal expression of p53 in Dgcr8−/− ESCs may lead to biological outcomes different from those elicited by full p53 activation upon genotoxic insult.

Activation of p53 regulates many biological processes, such as cell-cycle progression, cell viability, senescence, and metabolism (Kruse and Gu, 2009). All of these pathways have been shown to contribute to the tumor suppressor function of p53. Many of these known functions of p53 may also contribute to differentiation suppression. After NCS treatment, we observed that Dgcr8−/− ESCs are more prone to cell-cycle arrest compared with wild-type and Dgcr8−/−-302 ESCs (Figures 6B and 6C). Therefore, it is possible that increased cell-cycle arrest during differentiation could suppress tissue differentiation. Furthermore, it is possible that additional p53-mediated functions (e.g., apoptosis, senescence, etc.) are also differentially regulated, which might together contribute to the differentiation incompetence of Dgcr8−/− ESCs.

p53, referred to as the cellular gatekeeper or the guardian of the genome, is best known for its roles in responding to cellular stress and protection of genome integrity (Lane, 1992, Levine, 1997). Because Trp53 mouse embryos can develop into adulthood but succumb to tumors by 6 months of age (Donehower et al., 1992), p53 was thought to play a minimal role in embryo development. However, because most tumors arise in individuals later in life, it has been postulated that p53 must be evolutionarily selected by other non-tumor suppressor functions that are critical to earlier stages of life (Hu et al., 2008, Lu et al., 2009). Our data suggest that activation of p53 by a major genetic defect, such as miRNA deficiency, could serve as a potential mechanism to eliminate genetically defective embryos at very early stage of pregnancy to save maternal resources.

Experimental Procedures

Cell Culture and EB Differentiation

Mouse ESCs and iPSCs were passaged in mouse ESC maintenance medium (DMEM, 15% fetal bovine serum [FBS; Gemini Bio], 0.1 mM non-essential amino acid [NEAA, Thermo Fisher Scientific], 0.1 mM β-mercaptoethanol [β-ME, Sigma-Aldrich], and 1,000 U/mL mouse LIF [ESGRO, EMD Millipore]) on gelatin-coated tissue culture plates as described previously (Liu et al., 2015). For EB differentiation, trypsinized ESCs were suspended in Costar ultra-low-attachment cell culture plates (Corning) at a density of 1 × 105 cells/ml in EB medium (DMEM, 15% Knockout Serum Replacement [Thermo Fisher Scientific], 0.1 mM NEAA, 0.1 mM β-ME) for 4 days. The EBs were seeded onto gelatin-coated plates at a density of 5 EB/cm2 and continued in culture in EB medium for 14 days. For nutlin-3a treatment, ESCs were differentiated in EB medium with 10 μM nutlin-3a (Cayman Chemicals). For inhibition of TGF-β and/or BMP signaling, ESCs were differentiated in EB medium containing 15 μM SB431542 (Sigma-Aldrich) and/or 250 ng/mL Noggin (PeproTech). EB medium was changed every other day.

Vector Construction and Lentiviral Production

Sh-miRs and shRNAs were cloned into the lentiviral vector pLKO.1 at AgeI and EcoRI sites, as described previously (Moffat et al., 2006). DNA fragments containing the lLT antigen and T121 were PCR amplified from the pBABE-puro-SV40 LT plasmid (Addgene, no. 14088) (Zhao et al., 2003) and recombined into the pSINE-EF2-DEST-Pur lentiviral vector (Liu et al., 2015) using Gateway Technology (Thermo Fisher Scientific). Lentivirus was prepared as described previously (Zhao et al., 2014). Oligonucleotides used are listed in Table S4.

Immunostaining

Immunostaining was performed as described previously (Liu et al., 2017). In brief, EBs were fixed in 4% paraformaldehyde, blocked in Protein Block (Dako), and incubated with the appropriate primary antibodies overnight at 4°C and secondary antibodies for 1 hr at room temperature. Nuclei were counterstained by 0.5 μg/mL DAPI. Images were acquired by a Nikon Ti-S microscope and processed by Photoshop CS6. Antibodies used were TUJ1 (801202, BioLegend), MAP2 (sc-20172, Santa Cruz), NeuN (MAB377, Millipore), Synapsin (AB1543, Millipore), tyrosine hydroxylase (P40101, Pel-Freez), VGLU1 (135302, Synaptic Systems), and GABA (A2052, Sigma-Aldrich).

Mice, Teratoma Analysis, and iPSC Derivation

All animal experiments were performed in accordance with guidelines from the University of Alabama at Birmingham (UAB) and NIH. Teratoma injection was performed as described previously (Liu and Zhao, 2016). Non-obese diabetic severe combined immunodeficiency gamma mice (Jackson Laboratory) 4–10 weeks of age were injected subcutaneously with 1 × 106 to 5 × 106 ESCs. Tumors were harvested, fixed with 10% formalin, and processed by the UAB Comparative Pathology Laboratory. Tail tip fibroblasts (TTFs) were isolated from 2-week-old Dgcr8flox/flox;Trp53flox/flox mice (Marino et al., 2000, Wang et al., 2007) and cultured in D10 medium (DMEM, 10% FBS). Derivation of iPSCs was performed as described previously (Liu et al., 2015), which involves transduction of TTFs with STEMCCA lentivirus at MOI of 2, plating cells onto irradiated MEF feeders, and culturing in ESC maintenance medium for 3 weeks. Inactivation of Dgcr8 and Trp53 were achieved by transduction of a Cre-expressing retrovirus (Addgene, no. 24064).

Cell-Cycle and Apoptosis Analysis

BrdU pulse labeling and cell-cycle analysis were performed as described previously (Zhao et al., 2014). In brief, cells at 50% confluency were pulse-labeled with 10 μM BrdU (Sigma-Aldrich) 30 min before trypsinization and fixation in cold 70% ethanol at −20°C overnight. Cells were washed twice in PBS, denatured in 2N HCl containing 1% Triton X-100 at room temperature for 30 min, neutralized by 0.1 M sodium borate (pH 8.5), washed and resuspended in PBS containing 1% BSA and 0.5% Tween 20, and stained by APC-conjugated BrdU antibody (BioLegend) and 5 μg/mL propidium iodide. Cells were analyzed on a BD Fortessa flow cytometer. Apoptosis analysis was performed as described previously (Liu et al., 2017). In brief, ESCs were treated with the indicated concentration of nutlin-3a or 50 ng/mL NCS (Sigma-Aldrich) for 24 hr. Cells were trypsinized, stained with APC-conjugated Annexin V (BD Biosciences) on ice for 15 min and 1 μg/mL of propidium iodide (Sigma-Aldrich) at room temperature for 5 min, and then analyzed on a BD Fortessa flow cytometer. Data were analyzed by the FlowJo VX software.

RNA Extraction and qPCR

Total RNA was isolated with the DirectZol RNA Kit (Zymo Research) and cDNA were synthesized by the Verso cDNA Synthesis Kit (Thermo Fisher Scientific). qPCR was performed using 2× Absolute Blue Q-PCR Master Mix (Thermo Fisher Scientific) on a ViiA 7 real-time PCR system (Thermo Fisher Scientific). Primers are listed in Table S4.

Immunoblotting

Immunoblotting was performed as described previously (Liu et al., 2015). Whole-cell extracts were prepared in RIPA buffer (50 mM Tris-HCl [pH 8.0], 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, and 0.1% SDS), separated on a 4%–20% SDS-polyacrylamide gel (Bio-Rad), and transferred to polyvinylidene fluoride membrane (Thermo Fisher Scientific). For NCS treatment, 50 ng/mL NCS (Sigma-Aldrich) was added into medium and extracts were collected at the indicated times. Antibodies used were p53 (sc-6243) and glyceraldehyde-3-phosphate dehydrogenase (sc-25778, Santa Cruz).

Luciferase Assay

The full-length wild-type 3′ UTR of p53 was chemically synthesized and cloned into the pUC57 vector (Genewiz). The wild-type 3′ UTR of p53 was then subcloned into the pSiCheck2 plasmid (Promega) at XhoI and NotI sites to generate the WT-Luc reporter. Mutations at the miR-302 recognition site were introduced by PCR-based site-directed mutagenesis, as described previously (Zhao et al., 2014) to generate the Mut-Luc reporter. To generate the 302-Luc reporter, oligonucleotides complementary to miR-302d were ligated into the pSiCheck2 plasmid at XhoI and NotI sites. Oligonucleotides used for luciferase reporter construction are listed in Table S4. All constructs were confirmed by Sanger sequencing. Forty-eight hours before the luciferase assay, 3 × 103 HEK293T cells were transfected with 5 ng of luciferase reporters and 3 pmol of miR-302d mimic or control mimic conjugated with Dy547 (C-310372-05 and CP-004500-01, GE Dharmacon) by Lipofectamine 2000 (Thermo Fisher Scientific). Luciferase activity was measured using the Dual-Luciferase Reporter Assay System as per the manufacturer’s instructions (Promega) on a Synergy H1 Hybrid Multi-Mode Microplate Reader (BioTek).

Microarray, Small RNA-Seq, and Data Analysis

Global expression profiles were determined using the GeneChip Mouse Gene 2.0 ST Array (Affymetrix) at the Coriell Genotyping and Microarray Center. Raw microarray signal intensities were robust multiarray average summarized and quantile normalized using R/BioConductor (Bolstad et al., 2003, Gentleman et al., 2004, Irizarry et al., 2003). To detect differentially expressed genes, Student's t tests were performed on pairwise comparisons between Dgcr8-shctrl and Dgcr8-302 ESCs groups. We applied q value to t test p values to estimate the false discovery rate (Storey and Tibshirani, 2003), and set fold change larger than 1.5, and p < 0.05 as the threshold for determining the number of differentially expressed genes between groups. We used hierarchical clustering on the z score-transformed expression values of selected genes with the “complete” linkage method based on Euclidean distance to order the genes in the heatmap. Heatmap and scatterplot visualizations were performed in R (v.3.0.2) using the gplots and ggplot2 packages. GSEA analysis was performed according to Subramanian et al. (2005). Small RNA-seq was used to confirm sh-miR-302 expression. Total RNA was extracted from Dgcr8−/−-302 cells and submitted to the Genomic Services Lab at the HudsonAlpha Institute. The library was constructed by the standard miRNA library construction protocol (Illumina) and 15 million, 50 bp single-end reads were acquired. Adapters were removed from the reads using cutadapt (v.1.8.1) (Anders et al., 2015). All the reads were mapped to the mouse reference genome (GRCm38.74/mm10) using STAR aligner guided by a Gene Transfer File (Ensembl GTF version GRCm38.74) (Dobin et al., 2013).

Author Contributions

Z.L., K.K., and R.Z. conceived the experimental plan. Z.L., M.S., W.Z., D.K., C.-W.C., and J.F. performed experiments. C.Z., A.K.-J., and H.L. performed computational analysis. Z.L., X.H., T.M.T., H.L., K.K., and R.Z wrote the manuscript.