Author contributions: J.X.: collection and assembly of data, conception and design, manuscript writing, and final approval; S.B.H.L., M.Y.N., S.M.A., and J.P.K.: collection and assembly of data and final approval; V.L. and S.M.S.: data analysis and final approval; W.H.: conception and design, manuscript writing, and final approval.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS July 10, 2012.
ZO-1/Tjp1 is a cytosolic adaptor that links tight junction (TJ) transmembrane proteins to the actin cytoskeleton and has also been implicated in regulating cell proliferation and differentiation by interacting with transcriptional regulators and signaling proteins. To explore possible roles for ZO-1 in mouse embryonic stem cells (mESCs), we inactivated the ZO-1 locus by homologous recombination. The lack of ZO-1 was found to affect mESC self-renewal and differentiation in the presence of leukemia-inhibiting factor (LIF) and Bmp4 or following removal of the growth factors. Our data suggest that ZO-1 suppresses Stat3 and Smad1/5/8 activities and sustains extracellular-signal-regulated kinase (Erk) activity to promote mESC differentiation. Interestingly, Smad2, critical for human but not mESC self-renewal, was hyperactivated in ZO-1−/− mESCs and RhoA protein levels were concomitantly enhanced, suggesting attenuation of the noncanonical transforming growth factor β (Tgfβ)/Activin/Nodal pathway that mediates ubiquitination and degradation of RhoA via the TJ proteins Occludin, Par6, and Smurf1 and activation of the canonical Smad2-dependent pathway. Furthermore, Bmp4-induced differentiation of mESCs in the absence of LIF was suppressed in ZO-1−/− mESCs, but differentiation down the neural or cardiac lineages was not disturbed. These findings reveal novel roles for ZO-1 in mESC self-renewal, pluripotency, and differentiation by influencing several signaling networks that regulate these processes. Possible implications for the differing relevance of Smad2 in mESC and human ESC self-renewal and how ZO-1 may connect to the different pathways are discussed. Stem Cells2012;30:1885–1900
Tight junctions (TJs) are a hallmark of many differentiated cell types such as epithelia and endothelia, where they function both as barriers that regulate paracellular permeability and as fences that maintain the polarized distribution of plasma membrane components [1–3]. Many TJ constituents are also found in specialized cell-cell adhesion complexes of other cell types, for example, in cardiomyocyte intercalated discs or in Schwann cell myelin sheets . TJs are composed of integral membrane proteins such as Claudins, Occludin, and junctional adhesion molecules (JAMs), which are tethered to the actin cytoskeleton via scaffolding or adaptor proteins, in particular ZO-1/Tjp1, ZO-2/Tjp2, and ZO-3/Tjp3 . These adaptors also modulate cell proliferation and differentiation by binding and regulating the localization and/or function of transcription factors, transcriptional coactivators, and signaling proteins, often in response to cell-cell adhesion cues [5–7].
Using mouse preimplantation embryos, a relatively detailed understanding of the assembly of TJs and the role of Zonula Occludens proteins (ZOPs) in early mammalian embryogenesis has been obtained [8–10]. Activation of E-cadherin mediated adhesion between blastomeres at the eight-cell stage results in compaction and cell polarization [11, 12] and is required for the recruitment of ZO-1 to the apicolateral membrane contact sites of eight-cell blastomeres . At this stage, the ZO-1 α− isoform, which lacks the C-terminal α-exon, and JAM-1/JAM-A localize to E-cadherin-based adherence junctions (AJs) [13–15]. Asymmetric cell divisions of blastomeres at the 8- and 16-cell stages generate cells that will give rise to the trophectoderm epithelium, while the inwardly located daughter cells, devoid of epithelial characteristics, form the embryonic stem cells (ESCs) of the inner cell mass (ICM). Between the 16- and 32-cell stages, ZO-1 and JAM-1 segregate from the AJ in the developing trophectoderm to form TJs of characteristic morphology that contain the ZO-1 α+ isoform, ZO-2, and other components [14–16]. These TJs display the typical morphological characteristics and are functional in that they allow the vectorial fluid transport required for the formation of the blastocoel cavity [11, 17, 18]. In the late blastocyst, a second epithelium, the primitive endoderm, will arise from the ICM. When isolated, the ICM can recapitulate the differentiation of a trophectoderm-like epithelium and blastocoel formation [19–22]. Distinct roles for ZOPs in blastocyst morphogenesis have also been inferred from silencing ZO-1 and ZO-2 in mouse preimplantation embryos . In these studies, compensatory upregulation of ZO-1 was observed in ZO-2 knockdown embryos and correlated with a milder phenotype.
Recently, ZO-1, ZO-2, and ZO-3 have been inactivated in mice [23–26]. While ZO-1 and ZO-2 KO mice die during early embryonic development [24, 25], animals lacking ZO-3 show no overt phenotype [24, 26]. ZO-1-deficient embryos show massive apoptosis in the neural tube area, notochord and allantois, and defects in angiogenesis in the yolk sac , whereas embryos lacking ZO-2 fail to gastrulate . These studies provide in vivo evidence for nonredundant roles for the three ZOPs in mammalian development.
While the role for ZOPs in TJ formation during mouse preimplantation embryonic development has been characterized in some detail [8, 16], little is known about their role in differentiation from the ICM. ESCs are cell lines derived from the ICM that under appropriate conditions in culture can self-renew, retain their pluripotency, can differentiate into most if not all cell lineages, and are amenable to genetic manipulation [27–30]. Self-renewal and pluripotency of mouse ESCs (mESCs) depend on the balance of a combination of signaling pathways and transcription factors [30–32]. The cytokine leukemia-inhibiting factor (LIF) and fetal calf serum or bone morphogenic proteins (BMPs) maintain self-renewal and pluripotency by activating Stat3 and Smad1/5/8, respectively [33, 34]. Suppression of autocrine fibroblast growth factor-4 (Fgf4) signaling through the canonical Ras-mitogen activated protein kinase (MAPK) pathway maintains expression of pluripotency markers [35–37]. In addition, the key transcription factors Oct4, its coregulator Sox2, and Nanog specify the transcriptional regulatory network of mESCs [28-32, 38-40]. The integrated balance of these regulatory networks is thus required for mESC self-renewal, maintenance of pluripotency, and repression of differentiation [28, 30-32].
ZO-1 expression has previously been reported in human ESCs . Since ZO-1 is present in blastomeres [9, 14], its presence in ESCs could simply be a remnant from the asymmetric cell division of blastomeres that gives rise to the ICM cells. Alternatively, ZO-1 could serve a specific function in ICM cells and ESCs. To explore the latter possibility, we generated mESCs lacking ZO-1 using homologous recombination. We find that mESCs lacking ZO-1 are more refractive to the spontaneous outgrowth of differentiated cells, with differentiation in response to LIF/Bmp4 withdrawal being delayed. In the absence of ZO-1, Stat3 and Smad1/5/8, important for propagation as undifferentiated mESCs, are hyperactivated and protein levels of the pluripotency markers Nanog and Oct4 are sustained. Conversely, mitogen-activated protein kinase kinase (Mek)/Erk activity, required for mESCs to exit self-renewal, is attenuated. In addition, Smad2, important for self-renewal in human ESCs, is also activated in ZO-1−/− mESCs. The concomitant increase in RhoA protein levels suggests a shift in the Tgfβ/Activin/Nodal signaling from the noncanonical pathway, which regulates RhoA stability via Occludin, Par6, and Smurf1, to the canonical pathway leading to Smad2 phosphorylatyion. Bmp4, which in the absence of LIF promotes differentiation of mESCs down the epidermal lineage, failed to exert this effect on mESCs lacking ZO-1. Thus, ZO-1 modulates mESCs by suppressing self-renewal and promoting differentiation. These findings provide evidence for novel roles for ZO-1 in mESCs to modulate pluripotency and differentiation signals.
MATERIALS AND METHODS
mESCs were maintained without feeder cells. For serum-free culture, ESCs were cultured in N2B27-defined media containing a 1:1 ratio of Neurobasal media to Dulbecco's modified Eagle's medium/F12 media, supplemented with N2 and B27 (Invitrogen, Singapore, www.invirogen.com), 1,000 U/ml LIF (Chemicon, Singapore, www.millipore.com), and 10 ng/ml Bmp4 (R&D Systems, Singapore, www.rndsystems.com), as described previously [29, 42]. mESCs were seeded in six-well plates at a density of 2.5 × 105 cells per well in media with fetal bovine serum (FBS) and LIF. The next day (day 0), the media were changed to serum-free culture media. Human recombinant Fgf4 (R&D Systems, Singapore, www.rndsystems.com) was used at 5 ng/ml. PD98059 (Calbiochem, Singapore, www.merckmillipore.com) was used at 25 μM and SU5402 (Calbiochem, Singapore, www.merckmillipore.com) at 5 μM.
Generation of ZO-1−/− ESCs
Genomic fragments containing the ZO-1 locus were isolated from a mouse 129Sv genomic library (Stratagene) and subcloned into pBluescript II KS+ (Stratagene, Singapore, www.genomics.aligent. com). A targeting vector with a LacZ Neo cassette flanked by short and long arms of 3.6 kb and 6.7 kb for ZO-1 was designed to replace the exon containing the first transcriptional ATG. The targeting vector was linearized and electroporated into W4 (129S6) mESCs (Taconic Transgenics, Albany, NY, USA, www.taconic. com), which were then selected with 250 μg/ml G418 (Calbiochem, Singapore, www.merckmillipore.com). After 7–9 days selection, the colonies were picked and screened by long-range PCR. In order to obtain ZO-1−/− ESCs, ZO-1+/− ESCs were cultured in a 10 cm dish and selected with 10 mg/ml G418. After 7–9 days of selection, drug-resistant colonies were picked. Half of the cells from each clone were expanded on feeder cells, the rest were used for PCR screening. Correct targeting in the ESC clones was confirmed by Southern blot and absence of ZO-1 protein expression was verified by Western blot. Six independent ZO-1−/− ESC clones from two independent ZO-1+/− ESC clones were obtained from the screening. All the experiments were carried out using two independent ZO-1−/− mESC clones derived from independent ZO-1+/− mESC clones. Wild-type (WT) and ZO-1+/− mESCs served as controls.
Southern Blot Analysis
Genomic DNA was extracted from G418-resistant ESC clones and completely digested with ScaI. Two probes were used to check for the WT and mutant alleles. A 544-bp ZO-1 probe, corresponding to the 5′ end of the right arm of the target vector, was used to detect the 10 kb band from the WT ZO-1 allele and the 14.8 kb band in the mutant allele. A 668-bp Neo probe, corresponding to the neomycin resistance gene sequence, was used to confirm the 14.8 kb band in the mutant allele. The probes were labeled using the DIG DNA Labeling kit (Roche, Singapore, www.roche-applied-science.com) and hybridization of the labeled probe was carried out according to the manufacturer's protocol. Immunological detection of the hybridized probes was carried out using the protocol from the DIG Luminescent Detection kit for Nucleic Acids (Roche, Singapore, www.roche-applied-science. com). The blot was then exposed to x-ray film (FujiFilm, Singapore, www.fujifilm.com) for 20 minutes at room temperature.
Monitoring Outgrowth of Differentiated Cells
mESCs were seeded in gelatine-coated six-well plates at a density of 2.5 × 105 cells per well in media with FBS and LIF. The next day (day 0), the media were changed to serum-free culture media. After the indicated days in culture, 5–10 fields were randomly chosen and images were taken with a Nikon microscope and colonies with or without outgrowth of differentiated cells were counted. The experiment was repeated with a second independent ZO-1−/− mESC clone and the data were combined. For counting colonies with Gata4-positive cells, mESCs were seeded on glass coverslips in 24-well plates at a density of 5 × 104 cells per well in media with FBS and LIF. The next day, the media were changed to serum-free culture media either with or without LIF and Bmp4. After the indicated days in culture, the cells were fixed and processed for immunofluorescence microscopy with a Gata4 antibody as described below. Five to ten fields were randomly taken with a Zeiss microscope and the colonies with or without Gata4-positive cells were counted. The experiment was repeated with a second independent ZO-1 KO clone and the data were combined.
Colony Formation Assay
mESCs were seeded in gelatin-coated six-well plates at a density of 2.5 × 105 cells per well in media with FBS and LIF. The next day, the media were changed to serum-free culture media with LIF and Bmp4. After the indicated days in culture, the mESCs were trypsinized and 500 cells were seeded in a gelatin-coated six-well plate. In order to avoid apoptosis, the 500 cells were cultured in media with FBS and LIF. Colony number was counted after 5 days culture.
Cell Proliferation Analysis
mESCs were seeded in gelatin-coated six-well plates at a density of 1 × 106 cells per well in media with FBS and LIF. After 2 days culture, the cells were trypsinized and counted. Cells (1 × 106) were reseeded in gelatine-coated six-well plates. The culture was repeated for nine passages. ZO-1−/− mESC numbers were normalized to the WT ESCs. For the overgrowth experiment, 1 × 105 ZO-1−/− mESCs were mixed with 9 × 105 WT mESCs and seeded in gelatine-coated six-well plates in media with FBS and LIF (P1 culture). After 2 days in culture, the cells were trypsinized and reseeded in gelatine-coated six-well plates at a density of 1 × 106 cells per well. This was repeated for nine passages and the cells at P1, P3, P5, P7, and P9 were collected, lysed, and analyzed by Western blot using antibodies to ZO-1 and neomycin, which are only expressed in the WT and ZO-1−/− mESCs, respectively.
mESCs cultured on coverslips for the periods of time indicated were fixed in 4% paraformaldehyde at room temperature for 10 minutes for immunostaining. Primary antibodies were diluted in blocking buffer (4% goat serum and 10% BSA in PBS) and applied for at least 1 hour at room temperature or overnight at 4°C. Secondary antibodies conjugated to appropriate Alexa fluorophores (Invitrogen) were diluted 1:200 in blocking buffer and applied for 1 hour at room temperature. Images were obtained with a Zeiss microscope and associated software (Zeiss, Singapore, www.zeiss.com.sg). Primary antibodies used were rabbit polyclonal antibodies against ZO-1, ZO-2 (Zymed, Life Technologies Singapore, www.invitrogen.com, Cat# 61-7300, 71-1400), Oct4 (Cell Signaling, Singapore, www.cellsignal.com, Cat# 2750S), Gata4 (Santa Cruz, Singapore, www.scbt.com, Cat# K2007) and rat polyclonal antibodies against E-cadherin (Zymed, Life Technologies Singapore, www.invitrogen.com, Cat# 18-0223) and Troma1 (Troma1-C, Developmental Studies Hybridoma Bank at the University of Iowa, USA, www.dshb.biology.uiowa.edu).
Alkaline Phosphatase Staining
ESCs were seeded in six-well plates and cultured for the periods of time indicated. The cells were washed with PBS and then fixed and stained with the alkaline phosphatase (AP) staining kit (Sigma, Singapore, www.sigmaaldrich.com) following the instructions provided in the kit. Images were obtained with a Nikon inverted microscope (Nikon, Singapore, www.nikon.com.sg).
Cell Lysates and Immunoblotting
ESCs were washed twice with PBS and lysed for 45 minutes with RIPA buffer (1% deoxycholate, 1% Triton X-100, 0.5% SDS, 50 mM Na2HPO4, 150 mM NaCl, and 2 mM EDTA) supplemented with proteinase inhibitor (Roche, Singapore, www.roche-applied-science.com) and sodium orthovanadate (Sigma, Singapore, www.sigmaaldrich.com) on ice. Lysates were centrifuged at 13,000 rpm in a microcentrifuge for 15 minutes at 4°C and the supernatants were collected and analyzed by SDS-PAGE, Western blot, and chemiluminescence. Primary antibodies used were rabbit polyclonal antibodies against Oct4, Nanog, Stat3, pStat3, c-Myc, Erk, pErk, pSmad1/5/8, Smad2, pSmad2, and RhoA (Cell Signaling, Singapore, www.cellsignal.com, Cat# 2750S, 3580S, 9132S, 9145S, 5605S, 9108S, 9101S, 9511S, 3103, 3104S, and 2117P, respectively), Dab2 (BD Transduction Laboratories, Singapore, www.biomed.com.sg, Cat# 610464), Laminin1/2 (Abcam, UK, www.abcam.uk, Cat# ab7463), and rat polyclonal antibodies against Troma1 (Troma1-C, Developmental Studies Hybridoma Bank at the University of Iowa, USA, www.dshb.biology. uiowa.edu). For quantitation, films were scanned with a GS-800 calibrated densitometer (Bio-Rad, Singapore, www.bio-rad.com) and the densitometry readings were first normalized to the readings of the corresponding loading controls (GAPDH; glyceraldehyde-3-phosphate dehydrogenase) and then to the WT control. Data from at least three independent experiments were used for statistical analysis (Student's t test) using Prism (GraphPad Software Inc., La Jolla, CA, www.graphpad.com).
Cardiac and Neural Differentiation
Differentiation into the cardiac lineage was carried out as described . Briefly, 1,000 mESCs in a drop of 20 μl ES culture media with FBS without LIF were suspended under the cover of a bacterial culture dish for 2 days to form embryoid bodies (EBs). The EBs were then flushed into the culture media and cultured in suspension for another 5 days. On day 7, the EBs were seeded on coverslips in a 24-well plate and after another 5 days in culture, the cells were fixed and stained with cardiac-specific antibodies.
For neural differentiation, 5 × 104 cells per well were seeded on coverslips in a 24-well plate in media with FBS and LIF, and the next day, the media were replaced with N2B27 media and the cells cultured for another 10 days . The cells were then fixed and stained with neuron-specific antibodies.
Generation of Chimeric Mice
To confirm the pluripotency of the WT mESCs, the cells were used to generate chimeric mice. Cells (2.5 × 105) per well were seeded in a gelatine-coated six-well plate in media with FBS and LIF. The next day, the media were replaced with N2B27 media supplemented with LIF and Bmp4. After 2 days, the cells were trypsinized and injected into blastocysts derived from Albino B6 mice to generate chimera.
RNA Preparation and Reverse Transcription- Polymerase Chain Reaction Analysis
RNA was extracted from mESCs using the RNAeasy extraction kit (QIAGEN, Valencia, CA, www.qiagen.com). Semiquantitative reverse transcription-polymerase chain reaction (RT-PCR) analysis was carried out with the one-step RT-PCR kit (QIAGEN). The reaction was performed with 10 ng of mRNA in a 25 μl volume. The primer sequences, expected fragment size, and the number of cycles and temperature, respectively, for the listed gene-specific probes were as follows: Oct4 (forward 5′-GAG TAT GAG GCT ACA GGG AC-3′; reverse 5′-AAT GAT GAG TGA CAG ACA GG-3′; 379 bp, 28 cycles, 55°C), Nanog (forward 5′-GGC AGC CCT GAT TCT TCT ACC A-3′; reverse 5′-CTC CTC AGG GCC CTT GTC AGC-3′; 340 bp; 30 cycles; 55°C), Bmp4 (forward 5′-CAT GCT AGT TTG ATA CCT GAG ACC-3′; reverse 5′-CGT GAT GGA AAC TCC TCA CAG TGT-3′; 304 bp; 31 cycles; 55°C), Activin (forward 5′-ATA GAG GAC GAC ATT GGC AGG AGG-3′; reverse 5′-GCT GCT GAA ATA GAC GGA TGG TGA-3′; 229 bp; 34 cycles; 55°C), Tgfβ1 (forward 5′-TGC AGA GCT GCG CTT GCA GAG ATT-3′; reverse; 224 bp; 30 cycles; 55°C), Fgf5 (forward 5′-GAT GGC AAA GTC AAT GGC TCC-3′; reverse 5′-TTC CTA CAA TCC CCT GAG ACA CAG-3′; 460 bp; 32 cycles; 55°C), and actin (forward 5′-GGC CCA GAG CAA GAG AGG TAT CC-3′; reverse 5′-ACG CAC GAT TTC CCT CTC AGC-3′; 460 bp; 25 cycles; 55°C).
To connect the analyzed genes and proteins into a pathway diagram, we used KEGG , GeneMANIA , and STRING  to identify network links between key pathway players. This was complemented with protein-protein interaction data from our in-house databases  and further literature searches.
ZO-1 Is Expressed in mESCs
Western blot and immunofluorescence experiments were carried out to determine whether ZO-1 is expressed in mESCs. As shown in Figure 1A, ZO-1 protein was detected by Western blot in proliferating mESCs and expression levels remained constant over the 6-day period monitored. By immunofluorescence microscopy, ZO-1 was detected in Oct4-positive mESC colonies, where it apparently localized to the plasma membrane region and did not overlap with the Oct4 staining (Fig. 1B).
Generation of mESCs Lacking ZO-1
The ZO-1 locus was targeted in W4 mESCs with a β-galactosidase gene (lacZ) knock-in targeting vector (Supporting Information Fig. S1A). Using this strategy, the lacZ gene is placed in-frame downstream of the initiation ATG of the ZO-1 gene, resulting in a null mutation of the corresponding gene. G418-resistant mESC clones were screened for homologous recombination at the ZO-1 locus by Southern blot hybridization using 5′-arm and neomycin-specific probes and Sca1-digested genomic DNA. As expected, the gene-specific probe hybridized with a 10 kb and a 14.8 kb fragment (Supporting Information Fig. S1A, S1B), corresponding to the WT and mutant allele, respectively, whereas the neomycin-specific probe only detected the mutant allele (data not shown). Two clones showing homologous recombination were analyzed by Western blot and showed reduced ZO-1 protein levels as compared to WT mESCs (Supporting Information Fig. S1C). To inactivate the second allele, the two independent clones with homologous recombination were exposed to higher G418 concentrations and resistant clones were rescreened for homologous recombination at the second allele (Supporting Information Fig. S1B). One clone each was selected for the subsequent studies and gave comparable results. The clones showed no detectable ZO-1 protein expression as assessed by Western blot analysis (Supporting Information Fig. S1C) or immunofluorescence microscopy (Fig. 1B).
In the mESCs lacking ZO-1, protein levels for ZO-2 and ZO-3 and other TJ markers analyzed were comparable to those in WT or heterozygous cells (Fig. 1C). Plasma membrane localization of ZO-2, as assessed by its colocalization with E-cadherin, was also unaffected (Fig. 1D).
ZO-1−/− mESC Colonies Show Reduced Outgrowth of Differentiated Cells
By light microscopy, ZO-1+/+ and ZO-1−/− mESCs cultured for 2 days in N2B27 media supplemented with LIF and Bmp4 had a similar appearance except that more rounded-up apoptotic cells (see below) were present in ZO-1 KO cultures (Fig. 2A). By day 6, an increasing number of differentiated cells (e.g., epithelial-like, E-cadherin-positive cells as described by Ying et al. ) grew out from the AP positive undifferentiated ZO-1+/+ but not the ZO-1−/− mESC colonies (Fig. 2A and Supporting Information Fig. S2A). Approximately 30% of the colonies in control cultures showed outgrowth of differentiated cells, compared to ∼ 2% in ZO-1−/− mESC cultures (Supporting Information Fig. S2B). Outgrowth of differentiated cells from ZO-1−/− colonies was also inhibited when cultured in FBS containing media supplemented with LIF (Supporting Information Fig. S2C). RT-PCR analysis of different lineage markers in ZO-1+/+ and ZO-1−/− mESC cultures showed no differences in the expression of ectoderm (Fgf4 and Fgf5) and mesoderm (Brachyury, Goosecoid, and Bmp4) markers but a reduction in the endoderm markers Gata4, Dab2, Troma1, and Laminin1/2 in ZO-1−/− mESCs (data not shown). Immunofluorescence microscopy confirmed the expression of the differentiation marker Troma1 in the Oct4-negative differentiated cell outgrowth present in the WT and absent in the ZO-1−/− mESC cultures at day 6 (Fig. 2B). Likewise, a second differentiation marker, Gata4, was found in Nanog-negative cell outgrowth in WT mESC cultures and absent in those lacking ZO-1 (Supporting Information Fig. S2D–S2F). Tunnel staining did not reveal apoptosis in the bulk of the Gata4-positive cells in WT cultures (Supporting Information Fig. S2G).
Outgrowth of differentiated cells was not restricted to W4 mESCs but also observed to a similar extent for Go Germline and TT2 mESCs (Supporting Information Fig. S2H). W4 mESCs expressed similar transcript levels of pluripotency markers as the other mESCs as determined by RT-PCR, showed no expression of the differentiation marker Fgf5, but removal of LIF and Bmp4 induced expression of Fgf5 (Supporting Information Fig. S2I). W4 ZO-1+/+ (and ZO-1−/−) mESCs formed colonies with similar morphology when maintained in long-term culture by serial passage (Supporting Information Fig. S2J). Furthermore, when WT W4 mESCs cultured in N2B27 media supplemented with LIF and Bmp4 were injected into blastocysts, they generated chimera with a high percentage Agouti coat color (Supporting Information Fig. S2K) and germline transmission, suggesting extensive contribution to all lineages. Taken together, these data confirm intact self-renewal and pluripotency potentials for the W4 mESCs cultured in N2B27 media supplemented with LIF and Bmp4.
Enhanced Apoptosis in ZO-1−/− mESC Cultures
To determine whether the absence of differentiated cell outgrowth in ZO-1−/− mESC cultures reflected enhanced programmed death of differentiated cells, apoptosis in ZO-1+/+ and ZO-1−/− mESC cultures was monitored by cleavage of Lamin B and poly ADP ribose polymerase (PARP) and Casapase-3 activation. As compared to ZO-1+/+, enhanced apoptosis was detected in days 2 and 4 cultures of ZO-1−/− mESCs (Fig. 2C). Interestingly, in N2B27 media supplemented with LIF but lacking Bmp4, no difference in cell death between control and ZO-1−/− mESCs was observed. However, despite the inhibition of apoptosis to levels comparable to control mESCs, outgrowth of differentiated cells remained suppressed in ZO-1−/− mESC cultures (17% vs. 4%; Fig. 2D and Supporting Information Fig. S2L). In addition, cell death in both ZO-1+/+ and ZO-1−/− mESC cultures was dependent on cell density (Supporting Information Fig. S2M, S2N). Despite lower levels of apoptosis in denser ZO-1−/− mESC cultures, which were comparable to those observed in controls (Supporting Information Fig. S2M), expression of the differentiation marker Dab2 remained repressed (Supporting Information Fig. S2M) and outgrowth of differentiated cells remained inhibited in ZO-1−/− mESC cultures (Supporting Information Fig. S2N). In colony formation assays, ZO-1−/− mESCs formed significantly more colonies than wild-type controls (Supporting Information Fig. S2O), consistent with an enhanced self-renewal potential. Furthermore, exogenous Fgf4 was able to induce efficient differentiated cell outgrowth from ZO-1 KO mESCs (see below). Taken together, these observations suggest that apoptosis, although enhanced in ZO-1−/− mESC cultures, was not the primary cause for the absence of outgrowth of differentiated cells.
We next determined by Western blot analysis whether mESCs lacking ZO-1 showed changes in the expression levels of pluripotency and/or differentiation markers at days 2 and 6 of culture. In ZO-1−/− mESCs, protein levels for Nanog were elevated at days 2 and 6, and those for Oct4 increased at day 6 as compared to WT controls (Fig. 2E, 2F). At day 2, protein levels for the differentiation markers Troma1, Dab2, and Laminin1/2 were comparable in controls and ZO-1−/− ESCs but strongly elevated in WT controls by day 6.
The Absence of ZO-1 Represses the Mek/Erk Pathway
Autocrine Fgf4 induces mESCs to exit self-renewal by signaling through the Fgf-receptor (FgfR)/Grb2/Mek/Erk pathway and repression of Nanog expression [35, 36]. The Mek inhibitor PD98059 effectively blocked the outgrowth of differentiated cells from control mESCs colonies (Fig. 3A and Supporting Information Fig. S3A), suggesting that PD58059 phenocopies the lack of ZO-1. Analysis of the phosphorylation status of Erk, the direct downstream target of Mek, revealed that Erk phosphorylation was suppressed in days 2 and 6 cultures of ZO-1−/− mESCs as compared to controls (Fig. 3B, 3C). Surprisingly, addition of exogenous Fgf4 to ZO-1−/− ESCs induced the outgrowth of differentiated cells from ZO-1−/− mESCs (Fig. 3D and Supporting Information Fig. S3B). As assessed by Western blot analysis (Fig. 3E, 3F) and immunofluorescence microscopy (Fig. 3G), the differentiated cells outgrown from ZO-1−/− mESCs in the presence of Fgf4 expressed the differentiation markers Troma1, Dab2, and Laminin1/2. Phospho-Erk protein levels were strongly elevated in Fgf4-treated ZO-1−/− mESCs as compared to nontreated controls (Fig. 3E, 3F). Fgf4-induced differentiation could be blocked by the Mek inhibitor PD98059 or the FgfR inhibitor Su5402 (Supporting Information Fig. S3C). There were no apparent differences in Oct4 or Nanog levels between Fgf4 treated or control ZO-1−/− mESCs (Fig. 3E, 3F).
The Absence of ZO-1 Enhances Stat3, Smad1/5/8, and Smad2 phosphorylation and C-Myc and RhoA Protein Levels
In addition to favoring mESCs to exit self-renewal, ZO-1 may also suppress propagation as undifferentiated mESCs. We therefore analyzed the effect of the lack of ZO-1 on the LIFR/gp130, the BMPR/Alk, and the Tgfβ/Activin/Nodal pathways by monitoring Stat3, Smad1/5/8, and Smad2 phosphorylation levels, respectively. As shown in Figure 4A–4C, phospho-Stat3, phospho-Smad1/5/8, and phospho-Smad2 levels were augmented in ZO-1−/− mESCs of days 2 and 6 cultures. Consistent with the enhanced Stat3 activation, protein levels of the Stat3 target gene c-Myc were increased in ZO-1−/− as compared to control mESCs (Fig. 4A, 4B). Enhanced phospho-Smad2 levels in ZO-1−/− mESCs (Fig. 4A, 4C) suggest activation of the canonical Tgfβ/Activin/Nodal pathway, which was reported to promote proliferation of mESCs . In contrast, Tgfβ-mediated degradation of RhoA via activation of Par6/Smurf1 ubiquitin ligase [49, 50] may be suppressed given the increased RhoA protein levels in ZO-1 KO mESCs (Fig. 4A, 4C).
To test whether ZO-1−/− mESCs showed enhanced cell proliferation, ZO-1+/+ and ZO-1−/− mESCs were subjected to long-term culture and cell numbers were determined after different passage numbers. Plotting the ratio of ZO-1−/−/ZO-1+/+ mESCs showed an increase over time, consistent with higher proliferation rates for ZO-1−/− mESCs (Fig. 4D). To confirm this finding, ZO-1+/+ and ZO-1−/− mESCs were mixed at a ratio of 9:1 and cocultured. After different number of passages, the fraction of ZO-1+/+ and ZO-1−/− mESCs in the culture was determined by detecting ZO-1 and neomycin, respectively, by Western blot analysis. As shown in Figure 4E, the fraction of ZO-1−/− mESCs in the total cell population increased with passage number, consistent with higher proliferation rates for mESCs lacking ZO-1.
Lack of ZO-1 Delays mESC Differentiation Upon LIF/Bmp4 Withdrawal
We next determined whether mESC differentiation following the removal of LIF/Bmp4 was affected in the absence of ZO-1. As shown in Figure 5A, at day 2 in LIF/Bmp4-free media, 35% of control mESC colonies showed outgrowth of differentiated cells, which was only observed for 5% of ZO-1−/− mESC colonies. By day 6, however, differentiation was observed in both cultures, but the number of ZO-1+/+ and ZO-1−/− mESC colonies with differentiated cells remained lower as compared to controls (80% vs. 65%; Fig. 5A and Supporting Information Fig. S5A). Consistent with the data from the colony outgrowth assay, Western blot analysis showed elevated Oct4 and Nanog levels in ZO-1−/− mESCs as compared to controls at day 2, and these decreased in both by day 6 (Fig. 5B, 5C). In parallel, expression levels of the differentiation markers Troma1 and Dab2 increased over time in mESCs lacking ZO-1. However, the Mek/Erk pathway remained repressed as evidenced by lower phospho-Erk levels in ZO-1−/− mESCs at day 6. The biochemical data were further corroborated by immunofluorescence microscopy. Colonies in ZO-1−/− mESC cultures with low levels of Nanog (Fig. 5D) or Oct4 (Supporting Information Fig. S5B) showed less abundant outgrowth of differentiated cells or the differentiated cells showed less and weaker Gata4 (Fig. 5D, 5E) or Troma1 (Supporting Information Fig. S5B) expression, suggesting that the lack of ZO-1 delays but does not block differentiation.
Bmp4-Induced Differentiation Is Suppressed in the Absence of ZO-1
In addition to its role in promoting self-renewal of mESCs in cooperation with LIF, Bmp4 suppresses neurogenesis and promotes differentiation down the epidermal lineage [34, 51, 52]. To determine whether this effect of Bmp4 was affected by the absence of ZO-1, we analyzed the differentiation of mESCs in response to LIF withdrawal and Bmp4 supplementation.
Withdrawal of LIF and Bmp4 had no effect on ZO-1 expression in control mESCs (Supporting Information Fig. S6A). The efficient Bmp4-induced outgrowth of differentiated cells in control mESCs cultures was abrogated in ZO-1−/− cultures (Fig. 6A), with less than 20% of the colonies showing differentiated cells as compared to more than 80% in controls (Supporting Information Fig. S6B). Although Bmp4 lead to a similar reduction in Nanog levels from day 3 to day 5 in control and ZO-1−/− mESC cultures, Oct4 levels were not only elevated at day 3 in ZO-1−/− mESCs but also remained high (Fig. 6B, 6C). This correlated with the suppression of the Bmp4-mediated activation of the Mek/Erk pathway as evidenced by lower phospho-Erk levels at days 3 and 5 in ZO-1−/− mESC cultures when compared with controls. A similar increase in the differentiation markers Dab2 and Laminin1/2 was observed between days 3 and 5, but expression of Troma1 was suppressed in ZO-1−/− mESC cultures (Fig. 6B, 6C). Consistent with the biochemical data, Bmp4-treated control mESC cultures showed outgrowth of Troma1 and E-cadherin-positive, Gata4-negative cells at day 3 (Fig. 6D, 6E). The outgrowth of differentiated cells was absent from ZO-1−/− mESCs, which showed enhanced Oct4 labeling intensity as compared to controls.
To test whether the lack of ZO-1 also affected differentiation into other lineages, control and ZO-1−/− mESCs were cultured for 10 days in N2B27 media without LIF/Bmp4 to drive differentiation into the neural lineage. Alternatively, mESCs were grow for 7 days in hanging droplet cultures followed by 5 days in media with FBS but lacking LIF to facilitate differentiation into the cardiac lineage. ZO-1−/− mESCs were able to differentiate into both lineages (Supporting Information Fig. S6C, S6D).
ZO-1 is already expressed in blastomeres of the developing mammalian embryo, where it is recruited to AJs during compaction at the eight-cell stage . Blastomeres then undergo asymmetric cell divisions to give rise to the trophectoderm epithelium, where functional TJs are first assembled [11, 17, 18], and the ICM. ZO-1 protein is inherited by prospective ICM cells, but expression is lost in the ICM while ZO-1 transcripts are retained . Here, we show that mESCs, which are derived from the early preimplantation embryo ICM, express ZO-1. To explore possible roles for ZO-1 in mESC biology, we generated ZO-1 KO mESCs and found that ZO-1 modulates several signaling pathways involved in self-renewal, pluripotency, and differentiation. The enhanced proliferation of ZO-1−/− mESCs and the suppressed outgrowth of differentiated cells may thus reflect a status closer to cells of the ICM. However, ZOPs are also found in E-cadherin-based adherens type junctions  as well as gap junctions , which are both present in mESCs [55, 56]. Inactivation of ZO-1 in mESCs does not significantly upregulate expression of ZO-2 or ZO-3, indicating that these genes do not compensate for the lack of ZO-1. Furthermore, membrane localization of ZO-2 is not altered in ZO-1−/− ESCs, suggesting the interaction between ZO-1 and ZO-2  is not responsible for membrane recruitment of ZO-2 in mESCs.
ZOPs have best been characterized as components of TJs. While human ESCs display an epithelial-like polarized phenotype with electron dense, functional TJs , it is not known whether TJs are assembled in the ICM of the mouse embryo or in mESCs. Interestingly, disruption of polarity by matrigel or laminin overlay results in rapidly proliferating, pluripotent cells and it is not known whether these correspond to different pluripotency states . Laminin has also been shown to direct human blastomere differentiation to ICM . In the mouse early embryo, ZO-1 is inherited by prospective ICM cells but expression is then lost in the ICM while ZO-1 transcripts are retained . It is not clear whether the re-expression of ZO-1 in mESCs is a result of culturing or pluripotency state, but it indicates that mESCs and ICM cells are not identical.
Maintenance of the undifferentiated, pluripotent state of mESCs in culture depends on the presence of LIF and serum or BMPs, in particular Bmp4, and the key transcription factors Oct4 and Nanog [28-32, 38-40]. Binding of LIF and Bmp4 to their respective receptors results in the phosphorylation and activation of Stat3 and Smad1/5/8, respectively [29, 33]. Interestingly, the outgrowth of differentiated cells (e.g., epithelial-like, E-cadherin-positive cells as described by Ying et al. ) from mESC colonies observed in long-term (6-day) cultures is suppressed in mESCs lacking ZO-1. Outgrowth of differentiated cells is also observed in Go Germline and TT2 mESC cultures, both in the presence of LIF and serum or in defined medium supplemented with LIF and Bmp4, arguing against a cell type or culture condition-specific phenomenon. Furthermore, long-term culture of W4 mESC and a high contribution and germ line transmission in chimeric mice confirm intact self-renewal and pluripotency.
The reduced outgrowth of differentiated cells in ZO-1−/− mESC cultures could reflect enhanced apoptosis of differentiated cells, enhanced self-renewal of ZO-1−/− mESCs, a restricted differentiation potential, or a combination of these. While apoptosis is enhanced in ZO-1−/− mESC cultures, several lines of evidence indicate that this is not the primary cause for the absence of differentiated cell outgrowth. First, in the absence of Bmp4, apoptosis is comparable between control and ZO-1−/− mESCs cultures, yet outgrowth of differentiated cells remains suppressed in ZO-1 KO cultures. Second, the extent of apoptosis in ZO-1−/− mESC cultures is dependent on cell density. At densities where apoptosis is comparably low as in controls, expression of the differentiation marker Dab2 remains low and no differentiated cell outgrowth is observed. Third, Fgf4 rescues differentiated cell outgrowth in mESCs lacking ZO-1. Finally, ZO-1−/− mESCs form ∼ 60% more colonies in colony formation assays, suggesting enhanced self-renewal potential and consistent with the suppressed Fgf4 signaling and concomitantly enhanced LIF and Bmp4 signaling in ZO-1−/− mESC cultures.
Activation of the Mek/Erk pathway via autocrine Fgf4 induces mESCs to exit self-renewal [35–37]. Analysis of Mek/Erk pathway activation in ZO-1−/− mESCs revealed lower phospho-Erk levels (Fig. 7). Inhibition of Erk in control mESCs suppresses spontaneous differentiation, phenocopying the lack of ZO-1. Interestingly, exogenous Fgf4 induces Erk phosphorylation, differentiation marker expression, and outgrowth of differentiated cells. These findings suggest that the Fgf4 signaling pathway is functional in ZO-1−/− mESCs cells. Fgf4 expression, as assessed by RT-PCR, is also similar in control and ZO-1−/− mESCs (data not shown). One possibility is that autocrine Fgf4 levels are too low to initiate differentiation in mESCs cells if the Stat3 and Smad1/5/8 pathways are already enhanced, as is the case in ZO-1−/− mESCs. ZO-1 could alter the dose-response of Fgf4 on Erk signal transduction or influence autocrine secretion of Fgf4. ZO-1 could also modulate Erk activity through its direct interaction with Neph1 , a protein that modulates Erk signaling through its binding to Grb2 . Alternatively, ZO-1 could regulate Erk activity via its association with the Ras interacting protein AF-6  (Fig. 7).
The enhanced resistance of ZO-1−/− mESCs to differentiation furthermore correlates with an enhanced and sustained LIF and Bmp4 signaling, as evidenced by the hyperphosphorylation of Stat3 and Smad1/5/8, respectively (Fig. 7). Consistent with enhanced Stat3 phosphorylation, protein levels of the downstream target c-Myc  are elevated in ZO-1−/− mESCs, which also show higher proliferation rates, consistent with the hyperactivation of signaling pathways important for mESC self-renewal and maintenance of pluripotency. In contrast to mESC, which primarily rely on LIF/Bmp4 for self-renewal, the Tgfβ/Activin/Nodal pathway is critical for this process in human ESCs. Intriguingly, Smad2 is hyperphosphorylated in ZO-1−/− mESCs, indicative of enhanced Tgfβ/Activin/Nodal signaling activity (Fig. 7). While the exact role of Tgfβ/Activin/Nodal signaling pathway in self-renewal and differentiation in mESCs is still largely unclear, mESCs lacking E-Cadherin/β-catenin can use Activin/Nodal signaling via Smad2/3 to circumvent the LIF-dependent pathway to maintain self-renewal [56, 64, 65]. This pathway is also activated if E-cadherin homodimerization is inhibited and is probably regulated by specific domains in E-cadherin . Since ZO-1 can associate with E-cadherin structures in nonepithelial  and in nonpolarized epithelial  cells, the lack of ZO-1 may mimic the activation of Tgfβ/Activin/Nodal in the absence of E-cadherin. As observed in ZO-1−/− mESCs, activation of Activin/Nodal signaling was reported to stimulate mESC proliferation  and this could explain the delay in differentiation of mESCs lacking ZO-1 upon removal of LIF and Bmp4. Alternatively, since noncanonical Tgfβ/Activin/Nodal signaling via TgfβRI/II, Occludin, PAR6, and Smurf1-mediated ubiquitinylation and degradation of RhoA may be suppressed in ZO-1−/− mESCs (Fig. 7), the lack of ZO-1 could divert Tgfβ/Activin/Nodal signaling from the noncanonical pathway into the canonical, Smad2 dependent, pathway (Fig. 7). Occludin is expressed in mESCs and directly interacts with ZO-1 . Inhibition of Smurf-1, which can also regulate BMP signaling by tagging Smad1/5/8 for degradation, could contribute to the enhanced Smad1/5/8 activity in ZO-1−/− mESCs  (Fig. 7). In addition, since RhoA can activate Stat3 , accumulation of RhoA could contribute to the enhanced Stat3 phosphorylation and proliferation observed in ZO-1−/− mESCs (Fig. 7).
ZO-1 interacts directly with the C-terminal PDZ-binding motifs of several Connexins [70–72] and thereby influences gap junction assembly, stability, and function . Gap junction-mediated communication is observed between cells within the ICM and, during early postimplantation development, the expression of specific Connexins becomes restricted to cells of the same lineage, suggesting the formation of lineage-specific communication compartments . Cx31, Cx43, and Cx45 proteins, all reported to associate with ZO-1 [70-72, 75], are expressed and establish functional gap junctions in mESCs [76–78]. Cx31 is required for differentiation down the trophoblast cell lineage during placental development  and Cx45 null mESC shows no phenotype in vitro . In contrast, knockdown of Cx43 in mESCs inhibits cell proliferation but does not affect cell survival [55, 76] and results in the downregulation of stem cell markers and the upregulation of differentiation markers . In ZO-1−/− mESCs, Oct4 is upregulated and differentiation marker expression downregulated, suggesting that the effect of ZO-1 on mESCs is not directly linked to Cx43. Consistent with this interpretation is the observation that ZO-1−/− (data not shown) but not Cx43 knockdown  mESCs can form EBs. Interestingly, however, Laminin111-mediated stimulation of mESC proliferation was recently attributed to RhoA-mediated Cx43 phosphorylation and dissociation of the Cx43/ZO-1 complex , which could be mimicked by the absence of ZO-1 in the ZO-1−/− mESC.
In the absence of LIF, Bmp4 inhibits neural differentiation and promotes differentiation down the epidermal lineage [42, 51, 52]. While differentiation of mESCs lacking ZO-1 down the neural or cardiac lineages is not affected, ZO-1 is required for efficient differentiation of mESCs down the visceral endoderm lineage. This could explain in part the yolk sac vascularization defects implicated in the embryonic lethality of ZO-1 KO mice  since visceral endoderm, while not essential for differentiation of endothelial cells, is critical for the development and organization of yolk sac blood vessels .
Our data show that ZO-1 facilitates differentiation of mESCs down the epidermal lineage by inhibiting signaling pathways involved in mESC self-renewal and pluripotency (e.g., Stat3 and Smad1/5/8) and enhancing pathways that mediate differentiation (e.g., Erk). In addition, mESCs lacking ZO-1 show an activation of the canonical Smad2-mediated Tgfβ/Activin/Nodal signaling pathway and a possible suppression of the noncanonical Tgfβ pathway leading to the degradation of RhoA via Par6/Smurf1. ZO-1, which directly associates with Occludin, may thus function as a switch in Tgfβ signaling and its absence may shift the transduction of the signal from the noncanonical to the canonical pathway. Smad2 activation is thought to be inert in mESCs but central for self-renewal of human ESCs. It will thus be of interest to determine whether in human ESCs activation of Smad2 coincides with an attenuation of the Occludin/PAR6/Smurf1-mediated RhoA degradation, and whether this correlates with differences in the localization and/or function of ZO-1 in mouse compared to human ESCs.
The work was supported by the Agency for Science, Technology and Research (A*STAR), Singapore. V.L. is supported through A*STAR JCO Grant JCOAG04_FG03_2009. S.M.S. is an adjunct faculty member at the School of Biological Sciences, Nanyang Technological University (NTU), Singapore. W.H. is an adjunct faculty member at the Department of Physiology and the and Cancer Science Institute, National University of Singapore.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.