Differential requirements for β-catenin during mouse development

The development of multicellular organisms is governed by complex signaling cascades, which lead to defined gene expression patterns to balance the proliferation, migration and differentiation of embryonic cells. The highly conserved canonical Wnt signaling pathway and cadherin-mediated cell adhesion at adherens junctions provide two examples of major importance throughout embryogenesis (Nelson and Nusse, 2004). Vital to both of these processes is β-catenin, encoded by Ctnnb1. In cells expressing classical cadherins (i.e. E-cadherin or N-cadherin), β-catenin binds to the intracellular part of these transmembrane proteins and interacts with β-catenin, forming a chain of proteins connecting the cell membrane with the actin filament network of the cell (Kemler, 1993). Extracellularly, cadherins of the same isotype connect neighboring cells, and these homophilic interactions are the basis for the cell sorting processes that take place in embryogenesis (Takeichi, 1988). Interrupting adherens junctions at any point in this chain of proteins severely weakens the adhesive bond among cells, leading to incorrect cell sorting and migration, compromising development.

Canonical Wnt signaling is important for patterning the early embryo, organogenesis and maintenance of tissue-specific stem cells (Reya and Clevers, 2005). The molecular mechanisms that lead to the activation of the canonical Wnt signaling pathway are complex, involving the interplay of a multitude of inhibitors, activators and their co-factors at the cell membrane, in the cytoplasm and in the nucleus. Briefly, in the absence of canonical Wnt signals, the so-called destruction complex sequentially phosphorylates serine/threonine residues at the N-terminus of β-catenin, priming the protein for subsequent polyubiquitylation and proteasomal degradation (Aberle et al., 1997). Upon binding of a canonical Wnt ligand to receptors of the Frizzled and LRP family, the destruction complex is sequestered to the cell membrane and disintegrated (Zeng et al., 2008). Consequently, β-catenin accumulates in the cytoplasm, translocates to the nucleus and, in cooperation with LEF/TCF transcription factors, controls the transcription of canonical Wnt target genes (Daniels and Weis, 2005). Embryos lacking β-catenin (Haegel et al., 1995; Huelsken et al., 2000) or expressing a canonical Wnt signaling-incompetent protein (Valenta et al., 2011) have gastrulation defects and are lethal in early postimplantation development. An arrest in development is also seen upon expression of a dominant-active form of β-catenin that is missing the phosphorylation sites important for degradation. This results in accumulating amounts of β-catenin protein, which leads to precocious activation of Wnt target genes and the uncoordinated premature generation of mesodermal cells (Kemler et al., 2004). Hence, tight control of the expression level as well as the activity of β-catenin is essential for gastrulation.

Interestingly, β-catenin^flox/− mice that express β-catenin only from one allele develop indistinguishably from wild-type mice. In this regard it would be interesting to determine the limits of the physiological range of β-catenin expression within which development progresses normally. Here, we generated a conditional Ctnnb1 allele making use of the ubiquitously active ROSA26 promoter (Zambrowicz et al., 1997), and analyzed the consequences of predetermined levels of β-catenin in vitro and in vivo. Using different Cre mouse lines, differential amounts of β-catenin in various morphogenetic processes before and after gastrulation are examined. We found that even low levels of β-catenin expression are sufficient to maintain cadherin-dependent adhesion. By contrast, the expression of developmental stage-specific Wnt target genes requires different amounts of β-catenin, providing a molecular link between β-catenin dosage and the extent of morphogenesis.

Full-length Ctnnb1 was PCR amplified from a pYX-Asc plasmid (IMAGE Consortium) and cloned into the unique XbaI site of the pROSA26-1 targeting vector (Soriano, 1999). The targeting vector also included a splice acceptor and a floxed lacZ-neomycin (β-geo) selection cassette upstream of Ctnnb1. The splice acceptor mediates the expression of β-geo or Ctnnb1, respectively, joined to the endogenous, otherwise non-coding ROSA26 transcript.

W4/S129S6 wild-type and β-cat^flox/− embryonic stem (ES) cells cultured under standard conditions (DMEM with 15% FCS, LIF, feeder cells) were electroporated with the linearized targeting vector. Cells were selected with 400 μg/ml G418. Surviving clones were PCR screened for homologous recombination and confirmed by Southern blot. Selected ES cells were transiently transfected with a Cre-IRES-GFP plasmid, sorted, and checked for Cre-mediated recombination by PCR (for primers see supplementary material Table S1) and Southern blot.

For stimulation of canonical Wnt signaling, ES cells were treated with 100 ng/ml recombinant mouse Wnt3a (R&D Systems) for 4 hours and directly lysed in Trizol reagent (Invitrogen). For embryoid body (EB) differentiation, 500 cells were allowed to form aggregates in hanging drops (30 μl) in ES cell culture medium without LIF for 2 days. EBs were cultured in suspension for 3 days and then plated in gelatin-coated wells. The medium was changed every 2 days. After 2 weeks, the EBs were assessed for contractile cells or fixed for immunofluorescent staining. For teratoma formation, 2×10⁶ cells suspended in 200 μl PBS were injected into the flanks of nude mice. Teratomas were collected after 4 weeks and analyzed.

Targeted wild-type ES cells were injected into C57BL/6 blastocysts. Chimeras were backcrossed to β-cat^flox/flox mice (Brault et al., 2001) on a C57BL/6 background. Compound β-cat^flox/flox;ROSA26::β-catenin mice were then crossed with either Zp3::Cre (de Vries et al., 2000), Sox2::Cre (Hayashi et al., 2002), Cdx1::Cre (Hierholzer and Kemler, 2009), Foxn1::Cre (Soza-Ried et al., 2008) or Wnt1::Cre (Danielian et al., 1998) mice. Detection of the vaginal plug was considered as embryonic day (E) 0.5. Embryos were isolated at various stages from E7.0 to E18.5 and treated according to the follow-up experiments.

For in vitro culture, E7.0 embryos were dissected in M2 medium (Sigma) and cultured in embryo culture medium (DMEM without Phenol Red, 8 mM l-glutamine, 2% penicillin/streptomycin, 2 mM sodium pyruvate, 0.002% β-mercaptoethanol, 2× non-essential amino acids, 20 μg/ml ascorbic acid) containing 50% heat-inactivated filtered rat serum (Millipore) for 12 hours with or without 200 ng/ml Wnt3a. Mice and embryos were genotyped with the primers listed in supplementary material Table S1.

For immunostaining, ES cells and 7 μm cryosections of embryos were fixed in 4% paraformaldehyde for 15 minutes on ice, and immunodetection was performed as described (Messerschmidt and Kemler, 2010). The following primary antibodies were used: Nanog (Messerschmidt and Kemler, 2010) at 1:50; Oct3/4 (monoclonal mouse, Santa Cruz Biotechnology) at 1:200; Sox2 (monoclonal mouse, Chemicon) at 1:500; GP84 (Vestweber and Kemler, 1984) at 1:200; vimentin (monoclonal mouse, Exbio) at 1:200; N-cadherin (monoclonal mouse, BD Biosciences) at 1:100; β-tubulin 3 (mouse, Sigma) at 1:500; NF160 (monoclonal mouse, Abcam) at 1:200; Pax6 (polyclonal rabbit, Covance) at 1:300; β-catenin (rabbit polyclonal, Cell Signaling) at 1:200; Gata4 (polyclonal rabbit, Santa Cruz Biotechnology) at 1:100; and Troma-1 (Kemler et al., 1981) undiluted supernatant. Tagged secondary goat antibodies against mouse, rabbit and rat were: Alexa Fluor 488 or 594 (Molecular Probes) at 1:500. Nuclei were stained with 10 μM DAPI (Molecular Probes) added to the embedding medium.

For histological analysis, fixed embryos or teratomas were dehydrated, embedded in paraffin and sectioned at 7 μm. The sections were rehydrated and stained with Hematoxylin and Eosin. Immunohistochemical detection was performed on rehydrated sections with primary antibodies against E-cadherin, N-cadherin and nestin (mouse monoclonal, Chemicon) at 1:30. The signals were detected with the EnVision Plus System (Dako) using the DAB peroxidase substrate (Sigma). For analysis of skeletal phenotypes, E18.5 embryos were stained with Alcian Blue/Alizarin Red as described previously (Mallo and Brändlin, 1997). For lacZ staining, the embryos were treated as described (Stemmler et al., 2005).

ES cells were homogenized in lysis buffer and cell lysates were probed for β-catenin at 1:1000, E-cadherin (monoclonal mouse, BD Biosciences) at 1:3000 and Gapdh (monoclonal mouse, Calbiochem) at 1:25,000 as described (Kemler et al., 2004). Nuclear and cytoplasmic extracts were prepared with the ProteoJET Kit (Fermentas) according to the manufacturer's protocol. Upon lysis of nuclei, this fraction was further treated with benzonase nuclease (Sigma) to break up the chromatin.

For cell surface co-immunoprecipitation, ES cells were washed with PBS and incubated on ice in 50% culture medium in PBS supplemented with 10 μl/ml rabbit serum containing the GP84 antibody for 1 hour. After three washes with ice-cold PBS, the cells were lysed in 0.5% NP40 lysis buffer for 20 minutes on ice. Then, 500 ng cell lysate was incubated with 20 μl Dynabeads Protein G (Invitrogen) overnight at 4°C. The beads were washed three times in PBST (PBS containing 0.1% Tween 20) and bound proteins eluted by boiling in SDS loading buffer for 5 minutes.

Whole-mount in situ hybridization using digoxigenin-labeled riboprobes was performed as described (Correia and Conlon, 2001). For quantitative (q) RT-PCR, total RNA was isolated with Trizol (Invitrogen) following the manufacturer's instructions. Then, 500 ng RNA was reverse transcribed using oligo(dT) primers (Roche), diluted 1:25 and amplified in qPCR using Absolute qPCR ROX Mix (Thermo Scientific) in combination with the Mouse Universal Probe Library (Roche). Results were obtained from at least three experiments performed in triplicate using the ΔΔCT method. Primers and probes are listed in supplementary material Table S2.

Wild-type (wt) and β-catenin^flox/− (β-cat^flox/−) ES cells were targeted to generate mice and to characterize the ROSA26::β-catenin allele in vitro. In targeted β-cat^flox/− ES cells, transient expression of Cre recombinase results in the simultaneous deletion of the endogenous, floxed Ctnnb1 allele and the removal of the selection cassette from the ROSA26 locus, activating the ectopic expression of β-catenin (ROSA26^β/+ ES cells) (supplementary material Fig. S1). ROSA26^β/β ES cells were generated by a second round of gene targeting. ES cells expressing β-catenin only from the ROSA26 promoter maintained their characteristic growth morphology in culture (supplementary material Fig. S2A-D).

β-catenin mRNA levels in wt, β-cat^flox/−, ROSA26^β/β and ROSA26^β/+ ES cell lines were determined by qRT-PCR (Fig. 1A). As expected, β-cat^flox/− ES cells had ~50% β-catenin mRNA compared with wt ES cells. The mRNA levels in ROSA26^β/β or ROSA26^β/+ ES cells were further reduced to a half or a quarter, respectively, of that of β-cat^flox/− ES cells. From this it can be concluded that the ROSA26 promoter generates ~25% β-catenin transcripts compared with the endogenous Ctnnb1 promoter. To establish whether the differences in mRNA levels led to different amounts of β-catenin protein, western blots on serial dilutions of total cell lysates from all four ES cell types were performed (Fig. 1B). Levels of β-catenin protein closely mirrored the mRNA levels, as β-cat^flox/−, ROSA26^β/β and ROSA26^β/+ ES cells expressed ~50%, ~25% and ~12.5% relative to wt ES cells. Furthermore, subcellular fractionation revealed that β-catenin is present in the cytoplasm and in the nuclei of all four ES cell types (supplementary material Fig. S3A).

These ES cells with defined levels of β-catenin were studied for the dual function of β-catenin in cell adhesion and Wnt signaling. In immunofluorescence for β-catenin, ROSA26^β/+ ES cells showed clear membrane staining, indicating that the low amount of protein produced can complex with E-cadherin (cadherin 1) (supplementary material Fig. S2I-L). Cell surface immunoprecipitation experiments using an antibody specific for the extracellular part of E-cadherin [GP84 (Vestweber and Kemler, 1984)] revealed a similar association with β-catenin in wt and ROSA26^β/+ ES cells, demonstrating the correct assembly of the cadherin-catenin complex even in ES cells with the lowest amount of β-catenin (Fig. 1C). The Wnt signaling function was examined by stimulating the ES cells with Wnt3a and studying the activation of the canonical Wnt target genes Axin2, Cdx1 and brachyury (T) (Fig. 1D; supplementary material Fig. S3B,C). Without Wnt3a treatment, T is expressed at very low levels in all four ES cell lines. However, upon stimulation, clear differences became apparent. In wt and β-cat^flox/− ES cells, T expression was induced at least 11-fold and reached comparable levels in both ES cell types. By contrast, ROSA26^β/β ES cells only showed a 5-fold activation, whereas in ROSA26^β/+ ES cells no increase was detected.

Taken together, we see that β-catenin produced from the ROSA26 locus is functional in cadherin-mediated cell adhesion as well as in canonical Wnt signaling. However, owing to the reduced transcriptional activity of the ROSA26 promoter, the signaling function of β-catenin is almost absent in ROSA26^β/+ ES cells and is compromised in ROSA26^β/β ES cells.

To characterize ROSA26^β/+ and ROSA26^β/β ES cells in more detail, qRT-PCR analysis was performed for the pluripotency transcription factors Pou5f1 (Oct3/4), Nanog and Sox2, as well as for other stemness markers such as Klf4 and Tert, and comparable expression levels to β-cat^flox/− or wt ES cells were found (Fig. 2A-C; supplementary material Fig. S3D,E). In ROSA26^β/+ ES cells, nuclear localization of Oct3/4, Nanog and Sox2 was observed, similar to the other three ES cell types (Fig. 2D-F). Thus, despite the different amounts of β-catenin protein among the four ES cell types, the expression of pluripotency genes appears unaffected. From these results, we conclude that low levels of β-catenin are sufficient to maintain the ES cell phenotype.

In vitro differentiation potential was investigated using embryoid body formation in hanging drops and replating on gelatin-coated dishes, followed by immunofluorescent staining for differentiation markers (Fig. 2G-I). In general, all four ES cell lines underwent epithelial-mesenchymal transition with downregulation of E-cadherin and expression of the mesenchymal markers vimentin and N-cadherin (cadherin 2) (Fig. 2G,H). However, ROSA26^β/+ ES cells exhibited a notable bias toward differentiation into neurons, as shown by the abundance of cell aggregates positive for β-tubulin 3 (β-T3) and Nefm [neurofilament 160 (NF160)] (Fig. 2I). Contractile cardiomyocyte precursors were never detected in ES cells expressing β-catenin only from the ROSA26 locus, in contrast to controls, in which these structures were rather frequent. Thus, the in vitro differentiation of ES cells expressing low amounts of β-catenin suggested a shift toward neural differentiation and low mesodermal cell fate specification. Further support for this came from the analysis of teratomas. Whereas β-cat^flox/− ES cells differentiated into derivatives of all three germ layers (Fig. 3A), teratomas derived from ROSA26^β/+ ES cells were largely composed of neuroectodermal cell conglomerates (Fig. 3B,C). In immunohistochemistry of ROSA26^β/+ teratomas, E-cadherin expression was greatly reduced (arrows, Fig. 3D) and a marked increase of N-cadherin and nestin was observed (Fig. 3E,F).

In summary, keeping ES cells in an undifferentiated state does not require high levels of β-catenin. Furthermore, low β-catenin expression levels are sufficient to induce differentiation in vitro and in teratomas; however, the developmental fate of ES cells expressing β-catenin only from the ROSA26 locus is biased toward neural fates.

Next, we induced the switch in β-catenin expression from the endogenous to the ROSA26 locus during embryonic development, using several Cre transgenic mouse lines. In a first set of experiments, Cre recombinase expression was driven by the zona pellucida glycoprotein 3 promoter (Zp3::Cre) or by the promoter of the Sox2 transcription factor (Sox2::Cre), resulting in the deletion of endogenous Ctnnb1 before gastrulation (Fig. 4). Zp3::Cre recombination is restricted to the maternal genome, but recombined alleles are present in embryonic and extra-embryonic tissues (de Vries et al., 2000). Sox2::Cre, if paternally inherited, is only active in the epiblast starting at E5.5, leaving extra-embryonic tissues unaffected (Hayashi et al., 2002).

Sole expression of ROSA26^β/+ or ROSA26^β/β was able to ameliorate the β-catenin knockout phenotype (Fig. 4A). However, compared with β-cat^flox/− embryos (Fig. 4B), the formation of proper embryonic structures was incomplete, regardless of whether such embryos were generated with Zp3::Cre or Sox2::Cre (Fig. 4C,D). Morphologically, E8.5 mutant embryos had a sac-like morphology, exhibiting invaginations (Fig. 4C-H,K,L). The entire embryonic structure was composed of two layers, of which the outer layer that faces the yolk sac cavity was reminiscent of the visceral endoderm, whereas the inner layer had characteristics of a pseudostratified epithelium. Further marker analysis demonstrated that the inner layer expressed transcription factors of early neural progenitors, such as Sox2 and Pax6 (Fig. 4E,F), and showed enhanced expression of N-cadherin (Fig. 4G), whereas E-cadherin expression was very low (Fig. 4H).

To assess whether gastrulation took place in mutant embryos, expression of the early mesodermal marker T was analyzed at E7.5 (Fig. 4I,J). In contrast to controls, no T message was detected in ROSA26^β/+ or ROSA26^β/β embryos (Fig. 4J). From these results we conclude that the low amount of β-catenin in pregastrulation ROSA26^β/+ or ROSA26^β/β embryos is insufficient to establish the posterior-anterior Wnt signaling gradient, a prerequisite for gastrulation. This was further supported by the introduction of the Wnt reporter BATGal (Maretto et al., 2003) into ROSA26^β/β mutant embryos, which failed to become activated. To determine whether the Wnt reporter in ROSA26^β/β embryos could be force-activated, E7.0 embryos were cultured in vitro with or without recombinant Wnt3a for 12 hours (Fig. 4K,L). Without stimulation the BATGal reporter remained inactive (Fig. 4K). However, Wnt3a addition resulted in the activation of the reporter in a group of cells (Fig. 4L). The positioning of these BATGal-positive cells within the mutant embryos appeared to be random (inset in Fig. 4L), which was most likely due to the exposure of the entire embryo to the exogenous Wnt3a signal. Nevertheless, these experiments demonstrate that β-catenin originating from the ROSA26 locus is able to activate the Wnt reporter gene, and that the amount of β-catenin is critical. These results provide convincing evidence that at gastrulation, high expression levels of β-catenin are required for proper development.

In order to circumvent the defects during gastrulation, we made use of Cdx1::Cre, which at E7.5 initiates recombination in all three germ layers in the posterior half of the embryo caudal to the heart anlage (Hierholzer and Kemler, 2009). Interestingly, ROSA26^β/β and ROSA26^β/+ embryos generated with Cdx1::Cre exhibited remarkable differences in the formation of embryonic structures (Fig. 5A-F). At E9.5, ROSA26^β/+ embryos showed a truncated tail bud region (Fig. 5A) and caudal development was severely impaired, such that at E14.5 these embryos consist of a head attached to internal organs including lung, liver and intestine (Fig. 5D). The urogenital system and mesoderm-derived tissues making up the body wall (muscles, ribs and limbs) were highly underdeveloped or absent. By contrast, E9.5 ROSA26^β/β embryos had tail buds of normal length (Fig. 5B). However, most distally they showed an improperly folded neural tube that remained open as development progressed (arrowheads, Fig. 5B,E). At E14.5, ROSA26^β/β embryos had developed small kidneys (asterisk, Fig. 5E), adrenal glands and gonads, but lacked a tail, hind limbs and showed malformations including a persistently open cloaca (arrow, Fig. 5E). Skeletal preparations of E18.5 ROSA26^β/β embryos showed deformed shortened fused ribs, vertebrae and digits, and rarely also rudimentary bones of the upper hind limb (supplementary material Fig. S4). Although ROSA26^β/β embryos developed until birth, they did not survive. These results clearly indicate that the degree of caudal development is β-catenin dosage dependent. This was also seen using the BATGal reporter (Fig. 5A-C). Activation of this Wnt reporter was severely impaired in the caudal half of ROSA26^β/+ embryos (Fig. 5A). By contrast, ROSA26^β/β embryos showed a BATGal activity pattern comparable to that of controls (Fig. 5B,C). It is noteworthy that the BATGal reporter, which could not be activated before gastrulation (Fig. 4K), showed normal activity thereafter.

For further analysis, RNA was isolated from the tail bud region of E9.5 embryos and the expression levels of β-catenin determined by qRT-PCR; these were similar to those found in ES cells (supplementary material Fig. S5). To shed some light on how the different β-catenin levels influence caudal development, the expression profiles of Wnt target genes and genes important for the formation and maintenance of the tail bud and presomitic mesoderm were assessed. Generally, the expression levels of known and putative Wnt targets, such as Cdx1, Fgf8, T and Wnt3a, as well as Tcf1 and Lef1, were significantly reduced in ROSA26^β/+ mutant embryos (Fig. 6A-F). In ROSA26^β/β mutants, the expression of most of these genes reached at least 60% of that of the control group of embryos. Interestingly, intermediate or even elevated β-catenin expression did not further increase the transcription of target genes. Especially notable are the expression profiles of Wnt3a (Fig. 6D), which was expressed in ROSA26^β/β embryos at comparable levels to controls, and of Cdx1 (Fig. 6A), which remained at very low levels in both mutants. In contrast to this and to the expression profiles of Tcf1 and Lef1 (Fig. 6E,F), an increase in the expression of Tcf3 and Tcf4 was only observed for ROSA26^β/+ embryos (Fig. 6G,H). Based on these observations, it can be concluded that β-catenin-responsive genes are detected at very low levels in ROSA26^β/+ embryos and that doubling the β-catenin dosage is sufficient to induce a significant increase. Another important finding is that each gene seems to require a specific level of β-catenin in order to be expressed at the wild-type level.

To address whether the expression of ROSA26^β/+ or ROSA26^β/β can support the development and maintenance of specific organs or tissues at later stages of development, we made use of Wnt1::Cre and Foxn1::Cre transgenic mice. Wnt1::Cre-specific deletion of Ctnnb1 results in dramatic brain malformations and defective craniofacial development (Brault et al., 2001). Exclusive expression of ROSA26::β-catenin in these areas rescues brain development and the formation of neural crest-derived facial structures in a dosage-dependent manner (Fig. 7A-C). ROSA26^β/β mutant embryos were born, breathed, but died within the first hours, probably because of their inability to feed (Fig. 7D). In contrast to all other Cre mouse lines used in this study, expressing ROSA26::β-catenin with the help of Foxn1::Cre in the thymic anlage, epidermis and hair follicles (Soza-Ried et al., 2008), resulted in viable offspring (Fig. 7F,G). The knockout of β-catenin in the Foxn1 expression domain is neonatal lethal due to skin lesions (J. Swann, personal communication). Strikingly, even one ROSA26::β-catenin allele was sufficient to restore skin epithelial integrity. Although mutant ROSA26^β/+ and ROSA26^β/β pups are of reduced size at postnatal day (P) 14 compared with control littermates they recover with age. Again, a β-catenin dosage-dependent phenotype is detected for these mice. ROSA26^β/+ mice exhibit an irregular fur pattern, in which broad stripes of hair loss are followed by intervals of regrowth (Fig. 7F). ROSA26^β/β animals have continuous fur cover that is thinner than that of wt mice (Fig. 7G).

In summary, tissue-specific expression of ROSA26::β-catenin produces dosage-dependent phenotypes at embryonic and postnatal stages. Interestingly, for Foxn1::Cre, the expression of a single ROSA26::β-catenin allele is sufficient to substitute for endogenous Ctnnb1 in terms of skin integrity and viability of the animals; no immunological defects were found.

In the work presented here, we take advantage of the ROSA26 locus as a stable source for β-catenin transcription and were able to limit the amount of β-catenin to ~25% or ~12.5% relative to wild-type levels upon ablation of endogenous Ctnnb1. The ROSA26 locus was identified in a random retroviral gene-trapping approach (Zambrowicz et al., 1997), and is characterized by its ubiquitous and moderated expression throughout embryonic development and in adult tissues (Kisseberth et al., 1999). Based on the expression of the β-geo fusion gene used for selection of targeted ES cells, we detected similar expression levels in various tissues of E14.5 embryos by qRT-PCR (supplementary material Fig. S6A), as well as in X-Gal-stained sections (supplementary material Fig. S6B).

Monoallelic expression of ROSA26::β-catenin seems unable to cope with the dual demand for β-catenin that is necessary to supply adherens junction complexes and to initiate Wnt target gene transcription. In ROSA26^β/+ ES cells, β-catenin is predominantly found at the cell membrane in a complex with E-cadherin, as in wt ES cells. The strong association between these proteins even at low β-catenin concentrations can be explained by the nature of the assembly process of cadherin-catenin complexes (Ozawa et al., 1989). β-catenin co-translationally binds to the cytoplasmic domain of E-cadherin already in the endoplasmic reticulum, protecting the cadherin molecule from degradation and aiding in its transport to the cell membrane (Chen et al., 1999; Curtis et al., 2008). Furthermore, it was shown using recombinant proteins that E-cadherin is able to outcompete β-catenin from associations with Apc or Lef1 (Hülsken et al., 1994; Orsulic et al., 1999). Therefore, the inability to activate the transcription of canonical Wnt target genes in ROSA26^β/+ ES cells could be explained by insufficient amounts of free β-catenin due to the constant sequestration of newly synthesized β-catenin protein to E-cadherin.

Despite its high affinity for E-cadherin, we could still detect small amounts of β-catenin in nuclear extracts of ROSA26^β/+ ES cells. Nevertheless, canonical Wnt target genes were not induced in these ES cells upon stimulation of the Wnt signaling pathway. An explanation for this observation can be found in Weber's law, which states that a stimulus is always perceived with respect to its background level. Goentoro and Kirschner proposed that Weber's law applies to canonical Wnt signaling in developing Xenopus embryos (Goentoro and Kirschner, 2009). According to their hypothesis, the embryos respond to the fold-change value in β-catenin levels upon Wnt signaling and not to the absolute amount of β-catenin after Wnt induction. Therefore, within a certain range, the exact levels of β-catenin present in individual cells before the signaling event do not determine the signaling outcome, as long as the fold-changes are maintained. Thus, the β-catenin expression level of ROSA26^β/+ ES cells might generate insufficient fold-changes upon Wnt stimulation and thus Wnt target genes are not turned on. By contrast, in ROSA26^β/β ES cells the fold-change in β-catenin levels upon Wnt stimulation is sufficient to activate the transcription of Wnt target genes. However, their absolute expression level is markedly reduced compared with control ES cells, which indicates a more complex regulation of target gene expression than just the fold-change in β-catenin levels.

Absent or reduced signaling cues promote the differentiation of ROSA26^β/+ and ROSA26^β/β ES cells toward neural fates, which is in accordance with a hypothesis known as the default model of differentiation (Glinka et al., 1997; Hemmati-Brivanlou and Melton, 1997; Kamiya et al., 2011). Under culture conditions, ROSA26^β/+ ES cells exhibit normal ES cell morphology and express pluripotency-associated transcription factors similar to wt ES cells. These findings raise the question of how β-catenin contributes to maintain pluripotency and add new aspects to the controversial role of Wnt signaling in this process (Wray and Hartmann, 2012). Independent reports on the derivation of β-cat^−/− ES cell lines with normal pluripotency marker expression profiles (Lyashenko et al., 2011; Wray et al., 2011) support the idea that ES cells can be maintained without β-catenin. In this regard, it was shown that plakoglobin is able to substitute for β-catenin in adherens junctions, highlighting the importance of E-cadherin-mediated cell adhesion for pluripotency. E-cadherin was proposed to be at the core of a distinct signaling network that, parallel to Wnt signaling, promotes the ES cell state (Xu et al., 2010). Furthermore, during reprogramming, the forced expression of E-cadherin significantly enhances the yield of induced pluripotent stem cell colonies (Redmer et al., 2011). It is generally believed that β-catenin prevents the differentiation of ES cells through the inhibition of Tcf3-mediated repression of pluripotency-associated genes (Sokol, 2011). Alternatively, TCF-independent mechanisms have been proposed in which direct interactions of β-catenin with either Oct3/4 or Satb1 were shown to promote the expression of pluripotency markers (Savarese et al., 2009; Kelly et al., 2011). Therefore, the small amount of β-catenin produced from the ROSA26 locus suffices to sustain pluripotency either by supporting cadherin-mediated cell adhesion or through Wnt signaling-independent mechanisms.

During embryogenesis, the time point of ROSA26::β-catenin activation is critical for the developmental outcome of the embryos. When expressed before gastrulation, ROSA26^β/+ or ROSA26^β/β mutant embryos possess an epithelial structure, which represents an improvement compared with knockout embryos for which no embryonic organization is observed (Haegel et al., 1995). However, neither ROSA26^β/+ nor ROSA26^β/β is able to activate the BATGal reporter or endogenous Wnt target genes such as T. Given the crucial role of T in the induction of mesoderm (Arnold et al., 2000), mutant embryos transfate into neuroectoderm, analogous to the differentiation potential of ROSA26^β/+ or ROSA26^β/β ES cells. Similar phenotypes are also found in mice, in which canonical Wnt signaling is disrupted through the knockout of Wnt pathway components upstream of β-catenin (Liu et al., 1999; Hsieh et al., 2003; Biechele et al., 2011). In these knockout mice, β-catenin cannot accumulate due to its uninterrupted degradation; however, its adhesive function is unchanged. From this, we conclude that gastrulation requires a high level of β-catenin and that low β-catenin levels are able to sustain normal cell adhesion in vivo. Interestingly, the necessity for such high β-catenin levels seems not to exist in postgastrulation development. Instead, we observed a strong correlation between the β-catenin gene dosage, the degree of morphogenesis among ROSA26^β/+, ROSA26^β/β or control embryos, and the induction of β-catenin target genes.

From our gene expression data, it appears that Wnt3a is already expressed at ~50% in ROSA26^β/+ and at wild-type levels in ROSA26^β/β tail buds. Despite the normal expression level of Wnt3a in ROSA26^β/β embryos, their phenotype is highly reminiscent of the Wnt3a knockout (Takada et al., 1994), showing that Wnt3a alone is not sufficient to control tail bud development. Other factors known to be involved in the caudal extension of the embryo are Fgf8 and Cdx1 (van den Akker et al., 2002; Naiche et al., 2011). However, Fgf8 expression reaches only ~60% and Cdx1 expression remains below 30% in ROSA26^β/β embryos relative to controls. These findings highlight the need for proper expression of Fgf8 and Cdx1 in tail bud morphogenesis. Moreover, we also show that different Wnt target genes require different levels of β-catenin to be activated, adding another regulatory level to canonical Wnt signaling. A likely reason for the diminished transcription of Wnt target genes in ROSA26::β-catenin mutant embryos is the reduced expression levels of Tcf1 and Lef1. Since these transcription factors are required for the activation of canonical Wnt target gene transcription, their reduced expression levels might cause globally decreased expression of Wnt target genes. In support of this notion, a recent publication proposed an additional regulatory circuit of Wnt signaling wherein Tcf3-mediated repression of Lef1 transcription is alleviated upon stimulation of the pathway (Wu et al., 2012). In agreement with their hypothesis, we observe that, in ROSA26^β/+embryos, the expression level of Tcf3 is increased and those of Lef1 and Tcf1 are greatly reduced. Thus, at low expression levels, β-catenin seems unable to counteract Tcf3-mediated repression of Lef1 and Tcf1. However, in ROSA26^β/β embryos, the expression levels of Lef1 and Tcf1 are partially restored. Based on these findings, we reason that canonical Wnt signaling activity itself depends on the amount of β-catenin. Once β-catenin expression is at least as high as that in β-cat^flox/− mice, adequate Wnt signaling activity is restored. The expression levels of all LEF/TCFs and that of most target genes remain unchanged, even in mice that have an elevated expression level of β-catenin (β-cat^flox/+ plus ROSA26^β/β or ROSA26^β/+). Thus, we conclude that the Wnt signaling machinery is able to cope with small increases in β-catenin expression.

Tissue-specific expression of ROSA26::β-catenin also shows β-catenin dosage-dependent phenotypes. However, different tissues seem to require different levels of β-catenin for their morphogenesis and functionality. For example, expression of ROSA26^β/β using Wnt1::Cre supports the formation of midbrain and craniofacial structures, although mutant pups die within a few hours after birth. Strikingly, mice expressing ROSA26^β/+ in the Foxn1 expression domain develop to term and thrive normally. This shows that low β-catenin levels that merely support its adhesive function are sufficient to restore the integrity of the skin epithelial barrier in vivo. Nevertheless, ROSA26^β/+ mice show an abnormal hair cycle, which is less severe in ROSA26^β/β animals, probably owing to improved Wnt signaling.

By expressing β-catenin from the ROSA26 promoter we were able to separate the adhesive from the signaling function of β-catenin without introducing any additional mutations into the molecule. More importantly, we show that a given level of β-catenin controls particular morphogenetic events in development, which is also transferable to the induction of specific β-catenin target genes.

We thank Riana Vogt for excellent technical support, preparing reagents and cell culture; Benoît Kanzler for blastocyst injection; Caro Johner and Manfred Mellert from the animal facility; and Andreas Hierholzer, Ignacio del Valle and Daniel Messerschmidt for comments and discussions regarding the preparation of the manuscript.