Differential requirement for the dual functions of β-catenin in embryonic stem cell self-renewal and germ layer formation

Characterization of β-catenin deficient mESCs

mESCs were derived from β-catenin fl/fl (β-cat^fl/fl) ²⁴ blastocysts and individual feeder-free β-cat^fl/fl mESCs were established using LIF and serum. Several β-catenin deficient mESC subclones (referred to as β-cat^Δ/Δ) were established from one β-cat^fl/fl mESC line (NLβ-12) using adenoviral mediated Cre-recombinase and two were used for further characterization (Fig. 1a). The parental β-cat^fl/fl mESC line, NLβ-12, was fully pluripotent as it gave rise to chimeric mice and transmitted through the germline (see Supplementary Information, Fig. S1 online).

β-cat^Δ/Δ and β-cat^fl/fl mESCs showed an overall similar proliferation behaviour and colony appearance (Fig. 1b, c). β-cat^Δ/Δ colonies stained negative for β-catenin by immunocytochemistry, while E-cadherin and plakoglobin staining and distribution were normal (Fig. 1d). Interestingly, staining intensity of the latter was increased, as were protein levels (Fig. 1d, e), but transcript levels were unchanged (Fig. 1f). Thus, similar to F9 teratocarcinoma stem cells, plakoglobin substitutes for β-catenin cell-adhesion function in mESCs ²². This was confirmed by immunoprecipitation experiments, showing increased plakoglobin levels bound to E-cadherin in β-cat^Δ/Δ versus β-cat^fl/fl mESCs (Fig. 1g).

Plakoglobin has reportedly a weak in vitro signalling activity ^{25, 26}. Therefore, we analysed TCF/LEF-mediated gene expression using the sensitive lentiviral luciferase β-catenin activated reporter (BAR) system, with 12 multimerised TCF/LEF binding sites ²⁷. Without external stimulation minimal reporter activity was detected in control and β-cat^Δ/Δ mESCs (Fig. 1h). Upon stimulation with the GSK3 inhibitor BIO, recombinant Wnt3a (rWnt3a), or Wnt3a conditioned medium (Wnt3aCM) BAR-reporter activity was increased in control, but not detectable above ground levels in βcat^Δ/Δ mESCs (Fig. 1i, and data not shown). Likewise, Axin2 levels were increased in control, but not in β-cat^Δ/Δ mESCs (Fig. 1j). Stimulation had no effect on plakoglobin levels, nor did plakoglobin knock-down affect basal Axin2 levels (data not shown). Thus, our data suggest that the 2-fold increase in plakoglobin cannot substitute for β-catenin/TCF/LEF-mediated signalling activity.

Analysis of stem cell markers

No difference in Oct4, Nanog, Rex1 and Sox2 transcriptional (Fig. 2a), nor Oct4 and Nanog protein levels (Fig. 2b) were observed between β-cat^fl/fl and β-cat^Δ/Δ mESCs. FACS analysis showed no difference in SSEA-1 levels (data not shown). Furthermore, Stat3 phosphorylation kinetics was not altered (Fig. 2c). Late passages of β-cat^Δ/Δ and β-cat^fl/fl mESCs stained positive for the self-renewal marker alkaline phosphatase (AP) (Fig. 2d) and showed no differences in colony formation ability (data not shown). Concomitantly, transcriptomes analyses of β-cat^fl/fl and β-cat^Δ/ΔmESCs under standard conditions revealed significant differences (P<0.05) in only five genes besides β-catenin; with l7Rn6, H19, and Rhox5/Pem being elevated, and Trpv6 and Daam1 being decreased (see Supplementary Information, Table S1).

Overall, our data suggest that β-catenin activity is not required for mESC self-renewal in presence of LIF and serum. To further establish that β-catenin is dispensable for mESC self-renewal, we tested β-cat^Δ/Δ mESCs under serum-free conditions using the 2i+LIF system, inhibiting mitogen-activated kinase kinase (MEK) using PD0325901, and GSK3 using CHIR99021 ¹⁶. β-cat^Δ/Δ mESCs could readily be established in 2i+LIF from β-cat^fl/fl mESCs using adenovirally provided Cre-recombinase, but displayed a different morphology (Fig. 2e). Yet, Nanog, Oct4, Sox2 and Rex1 expression levels were indistinguishable from controls (Fig. 2f). β-cat^Δ/Δ mESCs formed colonies also under serum-free conditions in presence of PD0325901+LIF or CHIR99021+LIF, but not under PD0325901+CHIR99021 conditions (see Supplementary Information, Fig. S2 online). This establishes that β-catenin deficient mESCs are absolutely LIF dependent and that the GSK3 inhibitor, CHIR99021, has weak, β-catenin-independent effects. However, β-cat^Δ/Δ mESC propagation was only possible under serum-free conditions in presence of PD0325901+CHIR99021+LIF or PD0325901+LIF (data not shown).

Minor cell-adhesion defects in mESCs

β-cat^Δ/Δ and control mESCs colonies looked similar in serum+LIF (Fig. 1c). However, we noticed subtle differences and analysed colonies at the ultrastructural level by transmission electron microscopy (TEM). This revealed cell-adhesion defects in β-cat^Δ/Δ mESC colonies, with apparent regions lacking cell-cell contacts (Fig. 3a), suggesting that plakoglobin cannot fully substitute for β-catenin. To show that plakoglobin indeed substitutes in part for β-catenin cell-adhesion, we performed siRNA knock-down (Fig. 3b). Corresponding with transfection efficiency about 85% of plakoglobin siRNA treated β-cat^Δ/Δ mESC colonies displayed a differentiated morphology, with cells being more dispersed and flattened. In β-cat^fl/fl only 5% had a different morphological appearance, while control, lamin A/C siRNA treated cultures, showed no alterations (Fig. 3c; see Supplementary Information, Fig. S3b online). Concomitant to the morphological changes, the cell-adhesion molecules PECAM-1 and E-cadherin were lost from the membrane and their protein levels reduced upon plakoglobin siRNA treatment (Fig. 3b–d). Expression of self-renewal markers, Nanog, Oct4 and Sox2, was, however not altered; only proliferation was slightly decreased (see Supplementary Information, Fig. S3c online). This shows that in absence of β-catenin plakoglobin is an essential component of cell-adhesion in mESCs, mediating adhesion via adherens junctions and desmosomes, and supports previous findings that cell-adhesion is not essential for self-renewal maintenance ^{16, 17}. Furthermore, our data establish no requirement for plakoglobin in self-renewal maintenance. Next, we performed rescue experiments expressing different β-catenin versions, wild-type (βcat^WT), C-terminal truncated (β-cat^ΔC), and a point-mutated version (β-cat^M6) ²⁸ from the Rosa26 locus in β-cat^Δ/Δ mESCs, referred to in the following as β-cat^rescWT, βcat^rescΔC, and β-cat^rescM6. All three variants showed membrane localization (see Supplementary Information, Fig. S4a online) and restored cell-adhesion, as well as membranous localization of E-cadherin and PECAM-1 upon plakoglobin siRNA treatment (Fig. 3c, d). All three interacted with E-cadherin and their presence resulted in a slight reduction of plakoglobin (see Supplementary Information, Fig. S4b online). A slight further reduction was observed using mESCs stably expressing β-catenin isoforms under the CAG promoter yielding slightly increased levels of β-catenin variants (see Supplementary Information, Fig. S4c, d online), supporting the notion that the increased plakoglobin levels are due to protein stabilization via its recruitment to adherens junctions substituting for the loss of β-catenin.

Cell-adhesion defects during EB differentiation

The in vitro differentiation potential of β-cat^Δ/Δ mESCs was analysed using EB differentiation. This revealed cell-adhesion defects at day 5, reflected in an increase of single cells and associated with a size decrease of β-cat^Δ/Δ compared to control EBs (Fig. 4a). Controls formed cyst-like structures at day 9, while this did not occur with β-cat^Δ/Δ cells (Fig. 4a, and data not shown). Concurrent with the appearance of cell-adhesion defects, plakoglobin levels decreased from day 5 onwards in β-cat^Δ/Δ, but also in control EBs (Fig. 4b). The reduced plakoglobin levels were probably insufficient to compensate for lack of β-catenin, while in control EBs β-catenin steadily increased. EB differentiation proceeded normally in the different Rosa26 rescued mESCs (Fig. 4a). To confirm previous observations suggesting that the βcat^ΔC and β-cat^M6 variants are compromised in TCF/LEF transcriptional activity ²⁸, we generated the corresponding stable BAR reporter lines and activated canonical Wnt-signalling using CHIR99021. Surprisingly low reporter activity was observed in βcat^rescWT and no activity in β-cat^rescΔC or β-cat^rescM6 mESCs (Fig. 5a). Since Rosa26-transgene expression was relatively low and given that residual TOPFlash reporter activity had been observed previously ²⁸, we also assayed BAR reporter activity in the stable CAG-rescue mESCs. Here, the CAG-WT^resc cells showed a more pronounced response and residual activity was observed in CAG-M6^resc, but not in CAG-ΔC^resc mESCs (Fig. 5a). Accordingly, expression of the endogenous Wnt/β-catenin/TCF regulated genes, Axin2, Cdx1 and Brachyury, was not up-regulated in β-cat^Δ/Δ and βcat-CAG^rescΔC mESCs (Fig. 5b). Together, our data show that the β-catenin ΔC variant is defective in TCF/LEF mediated signalling.

Subsequent EB differentiation studies were performed using the Rosa26 βcat^rescWT and β-cat^rescΔC mESCs, since β-cat^M6 mESCs showed residual TCF/LEF- mediated activity and observed transgene-silencing in CAG-rescued mESCs during EB differentiation (data not shown),

Rescue of Endoderm and Neuronal Differentiation Defects

Genetic inactivation of β-catenin in mice results in an early developmental arrest around E7.0, with defects in anterior-posterior axis, mesoderm, definitive endoderm and ectoderm formation ^{19, 29}. Maternal β-catenin is provided until E4.5 and might therefore compensate for loss of zygotic β-catenin activity obscuring its absolute requirement in early mouse development ^{29, 30}. EB formation from mESCs allows differentiation of descendents of all three germ layers and resembles to a large degree normal postimplantation embryonic development ^{2, 3, 31, 32}. Thus, enabling us to examine the in vitro differentiation potential of β-cat^fl/fl, β-cat^Δ/Δ, β-cat^rescWT, and βcat^rescΔC mESCs. Expression profiles of the early ectodermal markers Nestin, Fgf5 and Pax6 were similar in β-cat^Δ/Δ and β-cat^fl/fl EBs during differentiation (Fig. 6a), suggesting that ectoderm induction occurs normally. Nevertheless, differentiation ofβ3-tubulin (TuJ) positive neurons was severely compromised in β-cat^Δ/Δ EBs (Fig. 6b). In contrast, TuJ positive neurons were readily detected in EBs derived from βcat^rescWT and signalling-defective β-cat^rescΔC mESCs (Fig. 6b). Neuronal differentiation in monolayer cultures revealed that β-cat^Δ/Δ mESCs have, in principle, the potential to differentiate into neurons, but compared to β-cat^fl/fl mESCs this is strongly reduced (Fig. 6c). Again differentiation of TuJ positive neurons was restored in β-cat^rescWT and signalling-defective β-cat^rescΔC cells (Fig. 6c).

Mesendodermal marker, Mixl1, expression was absent in β-cat^Δ/Δ EBs, but restored in β-cat^rescWT and signalling-defective β-cat^rescΔC EBs (Fig. 7a). Similarly, expression of the endodermal markers, Foxa2 and Gata6, was absent in β-cat^Δ/Δ EBs (Fig. 7a), but restored in β-cat^rescWT and β-cat^rescΔC EBs (Fig. 7a, see also Supplementary Information, Fig. S5 online). Loss and rescue of endoderm formation in EBs derived from β-cat^Δ/Δ and β-cat^rescWT and β-cat^rescΔC mESCs, respectively, was confirmed by immunofluorescent staining for the endodermal markers Gata4, Foxa2 and Cxcr4 (Fig. 7b, c). In β-cat^fl/fl EBs endodermal cells stained positive for Gata4, E-cadherin, and Cxcr4 and located to the outer region associated with a fibronectin-positive basement membrane (Fig. 7b). In β-cat^Δ/Δ EBs, Gata4 staining was severely reduced, staining for fibronectin and E-cadherin was very dispersed, and the Cxcr4 positive cells did not localize to the outer layer (Fig. 7b). In contrast, an outer layer with localized Gata4, E-cadherin and Cxcr4 staining was restored in β-cat^rescWT and β-cat^rescΔC EBs (Fig. 7b). The presence of Cxcr4 suggests that definitive endoderm is being formed ³³. Foxa2 positive cells were detected in dissociated EBs derived from β-cat^fl/fl, β-cat^rescWT and β-cat^rescΔC mESCs, but not in EBs derived from β-cat^Δ/Δ mESCs (Fig. 7c). Thus, our data suggest that TCF/LEF-mediated signalling activity is not absolutely required for endoderm formation. In addition, our data suggest that the cell-adhesion function of β-catenin promotes neuron formation in vitro. In agreement with the in vitro data, we found that in vivo β-cat^rescWT and β-cat^rescΔC mESCs show a higher chimeric embryonic contribution compared to β-cat^Δ/Δ mESCs and that they can in principle contribute to all three germ layers, but were preferentially found in endodermal and ectodermal layers (see Supplementary Information, Fig. S6 online).

mESC Establishment, Culture Conditions, Blastocyst injections

Individual day 3.5 old blastocysts isolated from β-cat^fl/fl females intercrossed with β-cat^fl/fl males were seeded on irradiated MEFs in 4-well plates in ESC medium (high glucose DMEM, 15% FBS, 2mM L-glutamine, 2mM penicillin/streptomycin, 1mM sodium pyruvate, 1x MEM non-essential amino acids, 0.1mM β-mercaptoethanol, 1000 U/ml LIF). After 5–7 days the inner cell mass was dissociated using 0.05% trypsin-EDTA and plated on MEFs. From these cultures colonies with ESC-like morphology were selected, eventually adapted for feeder-free conditions and further characterized. β-cat^Δ/Δ mESCs were established from the fully pluripotent NLβ-12 β-cat^fl/fl mESC line through infection with an Adeno-Cre virus (University of Iowa) at MOI 150 for 5×10⁴ mESCs. 24 hrs later individual mESCs were seeded in 96-wells and PCR genotyped (primers: 5′-AGAATCACGGTGACCTGGGTTAAA-3′, 5′-CAGACAGACAGCACCTTCAGCACTC-3′, 5′-CAGCCAAGGAGAGCAGGTGAGG-3′). Two mESC clones of each genotype, β-cat^fl/fl and β-cat^Δ/Δ, were further characterized. For serum-free cultures mESCs were cultured in N2B27 medium (StemCellSciences) supplemented with 1000 U/ml EsgroLIF (Chemicon), 3μM CHIR99021 and 1μM PD0325901 (both StemGent). Blastocyst injections were performed with the mESC clone NLβ-12 (on feeder, passage 18). The resulting high-grade chimera was crossed to C57Bl/6 and the offspring was analysed by PCR for germ line transmission of the β-catenin floxed allele. Blastocyst injections were also performed for β-cat^rescWT and β-cat^rescΔC mESCs clones stably marked with H2B-Gfp (Addgene plasmid 11680) ⁶⁸ injecting 10–15 mESCs per blastocyst. Embryos were recovered 6 days later, staged and visually examined for their grade of chimerism using a LSM 710 Spectral Confocal microscope (Zeiss).

Rosa26 and stable pCAGGS rescue mESCs and siRNA experiments

Targeting into the Rosa26 locus was performed using a modified Rosa26 targeting vector, pR26hygro (5′homology arm-SA-XbaI-SV40polyA-loxP-PGKhygro-loxP-3′ homology arm-PGKDTA) (kindly provided by A. Wutz) into which either wild-type full-length mouse β-catenin or N-terminal myc-tagged human versions of C-terminal truncated (ΔC) or M6 β-catenin (expression plasmids kindly provided by R. Grosschedl) had been inserted into the XbaI site by blunt-end cloning. For the Rosa26-CAG rescue lines myc-tagged wild-type and ΔC β-catenin were first cloned behind the CAG-promoter, fragments containing the CAG promoter and the β-catenin variants were then isolated and inserted into the XbaI site of pR26hygro by blunt-end cloning. Correctly targeted mESCs were identified by Southern blotting using Rosa26 5′ and 3′ external probes. For stable CAG-mESC lines, open reading frames were inserted by blunt-end cloning into the XhoI site of a modified pCAGGS vector containing a PGK neomycin selection cassette. Individual neomycin-resistant clones were established and tested by western for their β-catenin levels. For transient siRNA transfections mESCs were seeded onto 6-well plates at 2.5 × 10⁴ cells/well the day before Lipofectamin2000 (Invitrogen) mediated transfection with 150 nM plakoglobin (5′-ACACCUACGACUCGGGCAU-3′, 5′-CCAAGCUGCUCAACGAUGA-3′, 5 ′-GGAACUACAGCUACGAGAA-3′, 5 ′-UGAGUGUAGAUGACGUCAA-3′) o r LaminA/C (5′-UGACCAUGGUUGAGGACAA-3′) OnTarget siRNA mix (Dharmacon).

Generation of Stable BAR Reporter Lines and Luciferase Assays

Stable BAR (beta-catenin activated reporter) and fuBAR (found-unresponsive beta-catenin activated reporter) mESCs were established by infecting mESCs with pBARLS and pFuBARLS lentiviruses followed by puromycin selection for 7 days. For TOPFLASH/FOPFLASH luciferase reporter assays BAR and fuBAR mESCs were seeded onto 12-well plates at a concentration of 5×10⁴ cells/well. The Wnt pathway was activated the next day by treatment with 200 ng/ml rWnt3a (R&D Systems), 50% Wnt3aCM from stable Wnt3a producing L-cells, 2 μM BIO (Calbiochem) or 3 μM CHIR99021 respectively. Luciferase assays were performed in triplicates after 24 hrs using Dual-Glo luciferase assay (Promega) and the Synergy Fluorescence Plate Reader using at least two biological samples.

EB and neuron differentiation cultures

For EB differentiation, 1×10⁶ feeder independent mESCs were plated onto 15 cm low attachment plates (Corning) and cultured in mESC medium without LIF (differentiation medium). For the hanging drop method 1×10³ cells were plated per drop of differentiation medium and transferred after 60 hrs to low attachment plates (Corning). EBs were collected at indicated time points fixed in 4% paraformaldehyde and processed for immunocytochemical stainings or embedded in OCT (TissueTek) for cryosectioning. For high-density monolayer differentiation, 4×10⁴ cells/cm2 mESCs were plated on laminin-coated coverslips in N2B27 medium (StemCellSciences) and cultured for 14 days. Medium was renewed every second day. TuJ-positive cells were counted in random fields.

AP staining, Immunohistochemistry and TEM

AP staining was performed using the alkaline phosphatase kit according to manufactures instructions (Sigma Aldrich). Immunohistochemistry was performed upon mESCs or EBs seeded on gelatin covered 12 mm glass coverslips (Novodirect) after 20 min fixation with 4% paraformaldehyde, followed by 10 min permeabilization in PBS with 0.5% Triton-X and a 1 hr blocking step with PBS/2% BSA. For immunohistochemistry on cryosection, EBs from hanging drop cultures were fixed for 30 min in 4% paraformaldehyde, embedded in OCT and sectioned at 8 μm. Incubation with primary antibodies against Nanog, CXCR4 (both 1:100; Abcam), β-catenin (anti-C-terminus 1:250; BD Transduction; anti-N-terminus 1:200; Alexis Biochemicals), E-cadherin, plakoglobin, PECAM-1 (all 1:200; BD-Transduction), Oct3/4 (1:100), Foxa2 (1:100), TuJ (1:200), GATA-4 (1:500) (all from Santa Cruz), Fibronectin (1:200; Sigma) was performed for 1 hr at RT or oN at 4°C in PBS/2% BSA. Incubation with corresponding secondary anti-mouse, anti-rabbit or anti-goat antibodies coupled with Alexa488 or Alexa674 (1:1000; Molecular Probes) were performed for 1 hr at RT in the dark in PBS/2% BSA. To stain nuclei, samples were incubated for 5 min in 10 μg/ml DAPI (Sigma) and mounted with ProLong® Gold antifade reagent (Molecular Probes). Pictures were taken using the LSM 510 Meta/Axiovert200M confocal microscope. For TEM, mESCs were grown for 2 days on gelatinized 12 mm UV-sterilized Aclar coverslips, fixed in 2% glutaraldehyde for 1 hr, rinsed 3 times with PBS, treated with 2% OsO₄, rinsed with PBS, dehydrated 10 min each in a graded series of EtOH (40, 60, 80, 95, 100%). For embedding coverslips were incubated in a 100% EtOH/epoxy resin mixture (1:1) for 30 min at RT, followed by two 30 min incubations of the coverslips on 100% epoxy resin after which the coverslips were placed onto resin filled lids of Beem capsules and polymerized for 48 hrs at 60 °C. Resin blocks were sectioned on a Leica Ultracut UCT microtome. Sections were collected on metal grids and post-stained with Reynolds lead citrate (PbNO₃)₂ and uranyl acetate. Images were taken on a FEI Morgagni 268 electron microscope.

PCR analyses, Microarrays, Western blots and Co-Immunoprecipitation

RNA was isolated using RNeasy Kit (Qiagen). 1–2 μg RNA was used to prepare cDNAs using the SuperscriptII first strand cDNA synthesis system (Invitrogen). Amounts equivalent to 10–20 ng of the initial total RNA input were used as templates for real-time and RT-PCR analyses using gene primer pairs spanning exon-intron boundaries. Real-time PCR was performed using SYBR greenI nucleic acid gel stain (Roche) and TAKARA rTaq at least in duplicates. Products were analysed by agarose gel electrophoresis and melting curves. Values were calculated using the comparative ΔC(t) method and normalized to HPRT values, for relative expression levels displayed in histograms the values of control fl/fl mESCs were always set to 1. For semiquantitative RT-PCR mGAPDH and mHPRT were used as controls. All primers are listed in Supplementary Information, Table S2 online. In-house microarray platforms were used to compare total RNA from β-cat^fl/fl and β-cat^Δ/Δ mESCs. Cy3 and Cy5-labeled cRNA was prepared from 3–5 μg of total RNA by oligo-dT-primed reverse transcription. Activation/repression ratios for individual spots were calculated using GenePixPro5-chip software. Two biological replicas were analysed.

For western blots and immunoprecipitation (IP) analyses, protein was extracted from cultured mESCs or EBs. Protein concentrations were determined after Bradford (Biorad). For westerns, if not otherwise noted 10–20 μg extract was loaded per lane. For nuclear extracts cells were fractionated after ⁶⁹ or the NE-PER Kit (Pierce) was used for nuclear and cytoplasmic protein extractions. For IPs Gammabind Plus Sepharose Beads (Amersham) and 1–2 mg of pre-cleared lysate were used. Incubations with 2 μg and 10 μg antibodies against E-cadherin and α-catenin, respectively, were performed at 4°C oN. Primary antibodies were directed against β-catenin (1:2000; anti-C-terminus, BD Transduction; anti-N-terminus, Alexis Biochemicals), plakoglobin (1:6000; (BD-Transduction), E-cadherin (BD- Transduction), α-catenin (Chemicon), tubulin (1:6000; Sigma), Stat3, phospho (705Tyr)-Stat3, myc-tag (Cell Signaling) and proteasome 20S-C2 (Abcam), HP1α (Cell Signaling), and used at dilution 1:1000, if not otherwise noted.