TFCP2L1 represses multiple lineage commitment of mouse embryonic stem cells through MTA1 and LEF1

Embryonic stem cells (ESCs) are derived from the inner cell mass of blastocysts and have unlimited potential for self-renewal while maintaining the capacity to produce three germ layers (Huang et al., 2015). ESCs were first isolated from mice (Evans and Kaufman, 1981; Martin, 1981). They can be maintained in serum-containing medium supplemented with leukemia inhibitory factor (LIF) (Smith et al., 1988; Williams et al., 1988) or serum-free medium supplemented with the mitogen-activated protein kinase kinase (MEK) inhibitor PD0325091 (PD03) and the glycogen synthase kinase-3 beta (Gsk3β) inhibitor CHIR99021 (CHIR), hereafter referred to as two small molecule inhibitors (2i) (Ying et al., 2008). LIF supports self-renewal by activating STAT3 (Niwa et al., 1998), and CHIR and PD03 maintain self-renewal by inhibiting GSK3 and MEK, respectively (Ying et al., 2008). Notably, the combined use of any two of LIF, CHIR, and PD03 is able to maintain mouse ESC (mESC) self-renewal (Wray et al., 2010), indicating that they share common or cross-compensatory downstream targets in promoting mESC self-renewal. Some important convergent transcription factors have been identified and can partially recapitulate the self-renewal-promoting effect of LIF, CHIR or PD03 when overexpressed (Martello et al., 2013, 2012; Miyanari and Torres-Padilla, 2012; Pereira et al., 2006; Qiu et al., 2015; Ye et al., 2013, 2016; Yeo et al., 2014). However, how these genes promote mESC self-renewal is still largely unknown.

To gain greater insight into how these ESC-specific transcription factors function mechanistically in ESCs, we focused, in this study, on transcription factor CP2-like protein 1 (TFCP2L1), a common target of the LIF- and 2i-mediated signaling pathways (Martello et al., 2013; Ye et al., 2013). TFCP2L1 is critical for the maintenance and induction of the naïve pluripotency state (Martello et al., 2013; Qiu et al., 2015; Ye et al., 2013). Overexpression of Tfcp2l1 can not only reproduce the effect of LIF but is also able to replace the function of 2i when combined with Klf2 (Qiu et al., 2015; Ye et al., 2013). Strikingly, only knockdown of Tfcp2l1, but not other STAT3 targets, can impair the self-renewal ability rendered by STAT3 activation (Huang et al., 2015; Martello et al., 2013; Ye et al., 2013). Apart from its role in promoting mESC self-renewal, Tfcp2l1 is also sufficient to reprogram mouse epiblast-derived stem cells (mEpiSCs) to the naïve pluripotent state (Martello et al., 2013; Qiu et al., 2015; Ye et al., 2013). mEpiSCs are isolated from epiblasts of post-implantation mouse embryos and share many features with human ESCs (Brons et al., 2007; Tesar et al., 2007; Ye et al., 2014). However, how TFCP2L1 specifies the properties of the undifferentiated state unique to mESCs is underexplored. In this study, we provide a more detailed mechanism that describes the pivotal role of TFCP2L1 in mESCs.

Our previous research showed that overexpression of Tfcp2l1 can render mESCs independently of LIF, while knockdown of Tfcp2l1 induces differentiation (Ye et al., 2013). To systematically detect the differentiation status of Tfcp2l1 knockdown mESCs, we examined the expression of lineage-specific markers. Tfcp2l1 knockdown led to a corresponding decrease in the expression of pluripotency genes (Nanog, Oct4, Sox2 and Esrrb) (Fig. 1A), while resulting in upregulation of some endoderm (Sox17, Gata4 and Gata6), mesoderm [Brachyury (T) and Mixl1] and trophectoderm (Cdx2, Eomes and Elf5) markers (Fig. 1B). However, the ectoderm markers (Nestin, Sox1 and Fgf5) remained unchanged compared with scramble shRNA-treated control cells (Fig. 1B), implying that TFCP2L1 has a specific role in sustaining mESC pluripotency by suppressing endoderm, mesoderm and trophectoderm specification.

To confirm the effect of TFCP2L1 on the differentiation of germ layers, we carried out embryoid body formation assays to recapitulate early mouse embryo development by using cells that contain a doxycycline (Dox)-inducible Tfcp2l1 transgene (Ye et al., 2013). Dox-inducible Tfcp2l1 mESCs remain undifferentiated in the presence of Dox without LIF (Fig. 1C,D). We then analyzed Dox-inducible Tfcp2l1 embryoid bodies and found increased expression of the pluripotency markers Oct4, Sox2, Esrrb and Rex1 in Dox-treated embryoid bodies formed on day 6 compared with untreated cells (Fig. 1E), while the expression levels of endoderm (Gata6 and Gata4), mesoderm (T and Mixl1) and trophectoderm (Cdx2, Gata3 and Elf5) markers were decreased (Fig. 1E). As expected, the ectoderm markers (Sox1, Fgf5 and Pax6) remained unchanged (Fig. 1E). These data collectively suggest that elevated Tfcp2l1 expression can suppress the differentiation ability of mESCs into endoderm, mesoderm and trophectoderm lineages.

TFCP2L1 comprises 480 amino acids (aa) and has several distinct domains to exert different function. It depends on the CP2-like domain to bind DNA (Kim et al., 2016), while it regulates transcription through its N-terminus (Rodda et al., 2001). Besides, the ability of TFCP2L1 to form a hexameric complex resides in the sterile α-motif (SAM) domain (Kim et al., 2016). The C-terminus may take part in protein−protein interactions (Kokoszyńska et al., 2008). To better define which domain is indispensable for the ability of TFCP2L1 to maintain and induce naïve pluripotency, we generated PiggyBac (PB) transposon-system-mediated expression constructs encoding full-length (FL) and four different mutant TFCP2L1 proteins that lack the N-terminus (ΔNT), CP2-like domain (ΔCP2), SAM-like domain (ΔSAM) or C-terminus (ΔCT) (Fig. 2A). Notably, the deleted amino acid sequences in these mutants did not overlap. For example, the N-terminus and CP2-like domain share 32 aa (aa 21−52); we removed the 1−20 aa fragment in the N-terminus to generate Tfcp2l1^ΔNT because this protein sequence does not shown strong homology to other members of the CP2 family (Rodda et al., 2001). However, we deleted the 21−242 aa region to get Tfcp2l1^ΔCP2 because the CP2-like domain (aa 21−242) is conserved in all the CP2 family members(Kokoszyńska et al., 2008). These FLAG-tagged mutated forms of Tfcp2l1 were successfully overexpressed in 46C mESCs (Fig. 2B), and all TFCP2L1 mutant proteins are located in the nucleus (Fig. 2C). After being cultured in serum-containing medium without LIF for 8 days, Tfcp2l1^FL, Tfcp2l1^ΔSAM and Tfcp2l1^ΔCT cells retained the typical undifferentiated mESC morphology as well as an alkaline phosphatase (AP)-positive staining profile (Fig. 2D,E). Immunofluorescence assays also showed high expression levels of the pluripotency genes OCT4 and NANOG (Fig. 2F), while Tfcp2l1^ΔNT, Tfcp2l1^ΔCP2 and PB control transfectants differentiated (Fig. 2D-F), indicating that Tfpc2l1 relies on its N-terminus and CP2-like domain to maintain the undifferentiated state of mESCs. Next, in order to determine whether the N-terminus and the CP2-like domain are together sufficient to promote self-renewal, we overexpressed FLAG-tagged Tfcp2l1^NT+CP2 in mESCs (Fig. S1A,B), resulting in TFCP2L1^NT+CP2 protein being located in the cytoplasm and also in the nucleus (Fig. S1C). After being cultured without LIF for 8 days, many Tfcp2l1^NT+CP2 transfectants were AP-positive, albeit the self-renewal-promoting efficiency was lower than for TFCP2L1^FL transfectants (Fig. S1D,E). The N-terminus and CP2-like domain exert the self-renewal-promoting effect of TFCP2L1 possibly through binding to lineage-specific genes and by inhibiting differentiation because the transcription repression activity and the DNA-binding ability of TFCP2L1 were localized to these regions (Kim et al., 2016; Rodda et al., 2001). Similar effects were also observed upon overexpression of the Tfcp2l1 mutant in another mESC subline, J1 (Fig. S2A-C).

Overexpression of Tfcp2l1 in mESCs can prevent their differentiation into a primed pluripotent state (Fig. S3), and is also sufficient to induce naïve pluripotency in mEpiSCs (Martello et al., 2013; Ye et al., 2013). We next overexpressed different mutants of Tfcp2l1 in CD-1 mEpiSCs. After 12 days of culture in the presence of 2i and LIF, Tfcp2l1^ΔNT and PB transfectants died or differentiated (Fig. 2G,H). Surprisingly, Tfcp2l1^ΔCP2 mEpiSCs also gave rise to AP-positive colonies, but their number was much lower than that of AP-positive cells generated from Tfcp2l1^FL, Tfcp2l1^ΔSAM or Tfcp2l1^ΔCT mEpiSCs (Fig. 2H). This might be due to the reprograming conditions (presence of 2i and LIF), which reinforced the reprogramming activity of Tfcp2l1^ΔCP2. Collectively, these data indicate that the N-terminus and CP2-like domain are important for TFCP2L1 to induce naïve pluripotency.

As described above, Tfcp2l1^ΔNT and Tfcp2l1^ΔCP2 failed to maintain mESC self-renewal when cells were cultured without LIF (Fig. 2D-F). However, there was a lot of spontaneous differentiation in the Tfcp2l1^ΔNT and Tfcp2l1^ΔCP2 transfectants, even in the presence of LIF (Fig. 3A). These cells expressed lower mRNA levels of the pluripotency markers Nanog, Oct4, Esrrb and Rex1 than mESCs expressing Tfcp2l1^FL, Tfcp2l1^ΔSAM, Tfcp2l1^ΔCT or PB vector (Fig. 3B). This difference might be because Tfcp2l1^ΔNT and Tfcp2l1^ΔCP2 competitively inhibit the function of endogenous Tfcp2l1, which would be consistent with the results of Tfcp2l1 RNA interference (RNAi) (Fig. 1B). We then evaluated the mRNA expression of several lineage markers in these mutant-Tfcp2l1 cells. Interestingly, we observed the strongest induction of endoderm markers (Gata4 and Gata6) in Tfcp2l1^ΔCP2 mESCs, whereas Tfcp2l1^ΔNT dramatically upregulated mesoderm (T and Mixl1) and trophectoderm (Cdx2 and Gata3) markers compared with those in PB-transfected cells (Fig. 3C), meaning that the CP2-like domain is important for TFCP2L1 to repress endodermal specification, while the N-terminus of TFCP2L1 is essential to suppress that of the mesoderm and trophectoderm.

LEF1, a downstream factor of the Wnt/β-catenin signaling pathway, is closely associated with differentiation events in mESCs and with mammalian development (Chen et al., 2013; Galceran et al., 2001; He et al., 2008; Merrill et al., 2001; Zhou et al., 1995). Notably, the LIF/STAT3 signaling pathway significantly inhibits Lef1 expression (Ye et al., 2017). Tfcp2l1^FL being a main target of STAT3 (Martello et al., 2013), TFCP2L1^FL protein was also inhibiting Lef1 expression, whereas TFCP2L1^ΔNT protein greatly increased transcription of Lef1 when compared with the Lef1 expression level in PB-transfected cells in the presence of LIF (Fig. 3D). Accordingly, expression of Lef1 in Dox-inducible Tfcp2l1 mESCs decreased in a Dox dose-dependent manner (Fig. 3E), implying that LEF1 can bridge TFCP2L1 signaling to mesoderm and trophectoderm genes. To confirm this hypothesis, we overexpressed HA-tagged Lef1 in Dox-inducible Tfcp2l1 mESCs (Fig. 3F). We found that, on the one hand, overexpression of Lef1 impaired the ability of TFCP2L1 to repress T, Mixl1, Cdx2 and Elf5 expression (Fig. 3G). On the other hand, TFCP2L1^ΔNT failed to increase T, Mixl1, Cdx2 and Elf5 expression, and the ability to induce differentiation in Lef1 knockdown mESCs cultured in the presence of LIF was lost (Fig. 3H,I). These results suggest that the inhibitory function of TFCP2L1 is responsible for mesoderm and trophectoderm specification, partially, by repression of Lef1. This is unsurprising, as T has previously been shown to be a transcriptional target of Lef1 (Galceran et al., 2001). Lef1 regulates the stable maintenance of T expression during gastrulation of mouse embryos and in rat ESCs (Chen et al., 2013; Galceran et al., 2001). Likewise, Lef1 is also closely involved in the regulation of Cdx2 in mouse and rat ESCs (Chen et al., 2013; He et al., 2008). Further studies are needed to determine how TFCP2L1 regulates the expression of Lef1. We have previously reported that upregulation of Lef1 alone is also associated with the emergence of endodermal genes (Ye et al., 2017). However, TFCP2L1^ΔCP2 did not upregulate Lef1 transcription (Fig. 3D). Additionally, knockdown of Lef1 in Tfcp2l1^ΔNT overexpressing cells did not decrease the expression of endoderm markers (Gata4 and Gata6) (Fig. 3H), implying that TFCP2L1 does not suppresses endoderm specification through Lef1.

Finally, we wanted to elucidate how TFCP2L1 inhibits endoderm differentiation. Previously, van den Berg et al. had purified potential interaction partners of TFCP2L1, including many transcription factors and chromatin-remodeling proteins that were unknown to associate with the ESC network (van den Berg et al., 2010). We screened these candidates and focused on metastasis-associated (MTA) family genes because MTA1 knockdown leads to significantly enhanced expression of endoderm markers in mESCs (Liang et al., 2008). We established Tfcp2l1 overexpressing mESCs (Fig. 4A,B) and then infected them with lentivirus small hairpin RNA (shRNA) against the MTA1, MTA2 or MTA3 (Fig. 4C; Fig. S4A,B). Only knockdown of MTA1 substantially induced PB-Tfcp2l1 mESC differentiation, even in the presence of LIF (Fig. 4D; Fig. S4C). Expression of the pluripotency markers Oct4, Nanog, Esrrb and Rex1 was downregulated in MTA1 RNAi cells, while that of the endoderm makers Gata6 and Sox17 was enhanced (Fig. 4E,F). Similar results were obtained in the J1 cell line (Fig. S2D,F). MTA1 knockdown might release the expression of endoderm genes, because MTA1 functions as a transcription repressor in mESCs (Liang et al., 2008).

To validate the interaction between MTA1 and TFCP2L1 in mESCs, we transduced HA-tagged MTA1 (HA-MTA1) and FLAG-tagged Tfcp2l1 (FLAG-Tfcp2l1) into the same 46C mESCs and observed that the HA-tagged MTA1 and FLAG-tagged TFCP2L1 proteins co-immunoprecipitated with each other (Fig. 4G). To further test the influence of MTA1 on self-renewal, we overexpressed HA-tagged MTA1 (PB-MTA1) (Fig. 4H) and found that overexpression of MTA1 delayed mESC differentiation in response to removal of LIF (Fig. 4I). However, Tfcp2l1 shRNA-induced differentiation in PB-MTA1 cells, indicate that knockdown of Tfcp2l1 can also suppress the self-renewal response to MTA1 (Fig. 4J,K). Together, these results illustrate that MTA1 and TFCP2L1 bind to each other, and that TFCP2L1 inhibits the formation of endoderm through interaction with MTA1, and highlights the crucial role of MTA1 in mESC maintenance. Members of the MTA family of proteins are components of the nucleosome remodeling and the deacetylase (NuRD) complex and many members of the latter are also associated with endoderm development in ESCs (Kaji et al., 2007; Liang et al., 2008; Zhao et al., 2017). It will be of great interest to determine whether MTA1 and TFCP2L1 interact with other NuRD subunits when the specification of endoderm is suppressed.

Our findings present previously unrecognized facets of TFCP2L1 and, thus, provide new mechanistic insights into the role of TFCP2L1 in sustaining the undifferentiated state of ESCs. To understand how TFCP2L1 exerts its pluripotency-sustaining function will be beneficial to future basic research as well as to the safe utilization of ESCs.

46C mouse ES cells (mESCs) provided by Qi-Long Ying (University of Southern California), were cultured on 0.1% gelatin-coated dishes, J1 mESCs (Stem Cell Bank of Chinese Academy of Sciences) were plated onto MEFs at a density of 100 cells cm⁻², at 37°C in 5% CO₂. Medium for routine maintenance was Dulbecco's modified Eagle's medium (DMEM, Hyclone) supplemented with 10% FBS (ExCell Bio, Australia), 1× MEM non-essential amino acids (Invitrogen), 2 mM GlutaMax (Invitrogen), 0.1 mM β-mercaptoethanol (Sigma) and 1000 U/ml LIF (Millipore). For CD-1 mEpiSCs (Kim et al., 2013), cells were maintained in serum-containing medium supplemented with Activin A (10 ng/ml; Peprotech), basic fibroblast growth factor (bFGF, 10 ng/ml; Peprotech), and 4 µM IWR-1 (Sigma). For monolayer differentiation, 2000 single mESCs were plated into 0.1% gelatin-coated six-well plates and cultured in serum-containing medium without LIF for 8 days. 293FT cells, provided by Qi-Long Ying (University of Southern California Los Angeles, CA), were cultured in mESC medium without LIF.

Construction of the Tfcp2l1 and Lef1 overexpression and knockdown vectors was carried out as previously reported (Ye et al., 2013, 2017). Overlapping PCR was used to generate the Tfcp2l1 mutant. For RNA interference (RNAi) in mESCs, small hairpin RNA (shRNA) constructs were designed to target gene-specific regions of Tfcp2l1, Lef1 or MTA1, and then cloned into pLKO.1-TRC (AgeI and EcoRI sites). The target sequences of shRNA were as follows: Tfcp2l1 shRNA-1: 5′-GCTCTTCAATGCCATCAAAGG-3′; Tfcp2l1 shRNA-2: 5′-GCAGGAATGTGAGGCCAAAGA-3′; Lef1 shRNA-1: 5′-GCGACTTAGCCGACATCAAGT-3′; Lef1 shRNA-2: 5′-GCATACCGCACCCTGCGATCG-3′; MTA1 shRNA-1: 5′-GCAGGATTGAAGAGCTTAACA-3′; MTA1 shRNA-2: 5′-GGACATATTGGAAGAAATA-3′.

Cells were fixed in 4% paraformaldehyde for 2 min at room temperature, washed in PBS and incubated in the dark in AP-staining reagent (85L3R-1KT, Sigma) for 30 min at room temperature After washing twice with PBS, cells were visualized under a Leica DMI8 microscope (Leica, Germany).

For gene overexpression, cells were transfected with 2 μg of PiggyBac (PB) inserted with genes plus 2 μg transposase vector by using LTX (Lipofectamine^® LTX with Plus™ Reagent; catalog no. 15338100, Invitrogen, USA) according to the manufacturer's instructions. For knockdown experiments, pLKO.1-TRC-based lentiviral vectors and packaging plasmids pMD2.G and psPAX2 were co-transfected into 293FT cells using LTX. Supernatant was collected after 48 h and passed through a 0.45 μm filter (Millipore). mESCs were cultured in the viral supernatant in the presence of 8 µg/ml polybrene (Sigma) for 48 h. Selection was started after 48 h by adding 2 µg/ml of puromycin or 8 µg/ml of blasticidin.

For ESC-to-EpiSC differentiation, 46C mESCs (5×10⁴) were plated on six well plates pre-coated with FBS and cultured in serum medium supplemented with Activin A (10 ng/ml; Peprotech), bFGF (10 ng/ml; Peprotech) and 4 µM IWR-1 (Sigma). Cells were digested with 0.5 mg/ml collagenase IV and then passaged at 1:10. For reprogramming, 1×10⁵ transfectants were plated onto 0.1% gelatin-coated six-well plates and cultured in mESC medium supplemented with LIF and 2i [3 μM CHIR (SML1046, Sigma) and 1 μM PD03 (PZ0162, Sigma)]. The number of AP-positive clones was counted under a DMi8 microscope (Lecia, Germany).

i-Tfcp2l1 mESCs (1×10⁷) were cultured in suspension on ultra-low adhesion plates in standard mESC serum medium without LIF. Aggregates were allowed to grow for 6 days, and samples were lysed with TransZol Up (ER501-01, TransGen Biotech, China) for qRT-PCR analysis.

Cells were lysed in ice-cold RIPA cell buffer (P0013B, Beyotime Biotechnology, China) supplemented with Protease Inhibitor Cocktail (DI111-02, TransGen Biotech, China). Proteins were separated by 10% polyacrylamide gel electrophoresis (PAGE) and electrotransferred to a polyvinylidene difluoride (PVDF) membrane. Probing was performed with specific primary antibodies and horseradish peroxidase (HRP)-conjugated secondary antibodies. Primary antibodies used were against TFCP2L1 (SC-8635; Santa Cruz, 1:200), FLAG-tag (mouse monoclonal, F1804-200UG M2, Sigma, 1:2000), hemagglutinin (HA)-tag (rabbit monoclonal, catalogue no.: 3724S, Cell Signaling Technology, 1:2000) and α-tubulin (mouse monoclonal, catalogue no.: 32-2500, Invitrogen, 1:5000).

Cell extracts were prepared by using Nonidet P-40 lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5% Nonidet P-40 and protease inhibitors). The supernatant was collected and incubated with 10 µl of anti-FLAG or anti-HA affinity gel (450-FG or -HA, GNI, Japan) for 2 h at 4°C. The beads were then washed six times with lysis buffer and resuspended in 1×SDS sample buffer for western blotting analysis.

Total RNA was extracted by using the TRIzol Up Plus RNA Kit (TransGen Biotech, China). cDNA was synthesized from 1 µg of total RNA by using the TransScript All-in-One First-Strand cDNA Synthesis SuperMix for qPCR (One-Step gDNA Removal, TransGen Biotech, China) according to the manufacturer's instructions. Quantitative real-time PCR (qRT-PCR) was carried out by using Top Green qPCR SuperMix (TransGen Biotech, China) in a PikoReal Real-Time PCR machine (Thermo Scientific, USA). Target gene expression was normalized against expression of β-actin. The primers used are listed in Table S1.

Cells were fixed in 4% paraformaldehyde for 20 min and incubated at 37°C in blocking buffer (PBS containing 5% BSA and 0.2% Triton X-100). Cells were incubated in the presence of primary antibodies at 4°C overnight. After being washed three times with PBS, cells were then incubated with Alexa Fluor 488 (Invitrogen, 1:1000) secondary antibody for 1 h at 37°C. Nuclei were stained by using Hoechst 33342 dye (Invitrogen, 1:10,000). Primary antibodies against Oct4 (catalog no. SC-5279, Santa Cruz, 1:200) and Nanog (catalogue no. ab808692, Abcam, 1:200) were used.

All data are reported as the mean±s.d. Student's t-test was used to determine the significance of differences in comparisons. P<0.05 was considered to be statistically significant.