Tcf7l1 protects the anterior neural fold from adopting the neural crest fate

Neural crest (NC) cells are vertebrate-specific, multipotent, highly migratory cells that originate from the ectoderm of the neural plate border (NPB). NC cells possess the ability to differentiate into a wide range of cell types, including the bone and cartilage of the skull, sensory neurons, outflow tract of the heart, enteric ganglia and pigment cells. During embryogenesis, a combination of gene expression driven by Wnt, FGF and BMP pathways determines the future characteristics of the head ectoderm, transforming it into either placodes, neural plate, epidermis or NC cells (Garnett et al., 2012; Groves and LaBonne, 2014; Litsiou et al., 2005; Simoes-Costa and Bronner, 2015; Steventon et al., 2009). Wnt/β-catenin represents one of the most extensively studied signaling cascades. Binding of secreted Wnt proteins to their Frizzled receptors and their Lrp5/6 co-receptors initiates a complex cascade of intracellular events leading to the stabilization of β-catenin. Stabilized β-catenin accumulates in the cytoplasm and enters the nucleus, where it binds to the Tcf/Lef family of transcription factors, thereby regulating the expression of Wnt target genes (Clevers and Nusse, 2012; http://web.stanford.edu/group/nusselab/cgi-bin/wnt/).

Four Tcf/Lef genes have been identified in vertebrates: Lef1 (Travis et al., 1991), Tcf7 (alias Tcf1) (van de Wetering et al., 1991), Tcf7l1 and Tcf7l2 (alias Tcf3 and Tcf4) (Castrop et al., 1992). Tcf/Lefs share several characteristic protein features, such as the N-terminal β-catenin-binding motif, GRG motifs for binding of Groucho/TLE co-repressors, and the HMG box containing the DNA-binding domain. Tcf7 and Tcf7l2 also contain a C-clamp, a second DNA-binding motif that further modulates gene target specificity (Cadigan and Waterman, 2012). Distinct expression patterns of individual Tcf/Lefs and their unequal capability to activate/repress Wnt target genes represent the ultimate regulatory step of the Wnt/β-catenin signaling pathway. Tcf7l1, expressed in the anterior ectoderm at E7.5 and in the anterior neural fold (ANF) at E8.5 of the developing mouse embryo (Korinek et al., 1998; Merrill et al., 2004), is the least potent activator but the strongest repressor of the whole family (Cole et al., 2008; Liu et al., 2005; Ombrato et al., 2012). Tcf7l1-deficient mice, which display a duplicated axis and a severely reduced anterior part of the embryo, die shortly after gastrulation (Merrill et al., 2004). Tcf7l1 inactivation appears to allow Wnt/β-catenin signaling to spread towards the ANF in both zebrafish and mouse. This leads to a deficiency in the expression of ANF-specific genes, accompanied by severe defects in head formation (Andoniadou et al., 2011; Dorsky et al., 2003; Kim et al., 2000). All these data have been interpreted as consistent with the repression of Wnt signaling in the ANF being required for anterior-posterior (A-P) axis determination of the neural tube.

Experiments in Xenopus, zebrafish and chicken have shown a requirement for Wnt/β-catenin signaling from the earliest steps of NC development, at first during gastrulation when Wnt/β-catenin signaling drives the expression of the NC specifier genes Pax3 (Bang et al., 1999; Garnett et al., 2012; Taneyhill and Bronner-Fraser, 2005; Zhao et al., 2014), Foxd3 (Janssens et al., 2013) and Gbx2 (Li et al., 2009), and later during NC maintenance (Kléber et al., 2005; Steventon et al., 2009) and the epithelial-to-mesenchymal transition (EMT), via triggering Slug (Snai2, or Snail2) gene expression (Burstyn-Cohen et al., 2004; Sakai et al., 2005; Vallin et al., 2001). In contrast to non-mammalian experimental models, a clear picture of the role of Wnt/β-catenin signaling in early NC cell populations in mammals is absent. This is partly caused by a lack of appropriate Cre drivers for the early pre-migratory NC lineage. The widely used Wnt1-Cre mouse targets the NC population after its induction and specification (Danielian et al., 1998; Lewis et al., 2013). Depletion of β-catenin using Wnt1-Cre leads to malformations of the craniofacial structures, accompanied with a deficiency in sensory neurons (Brault et al., 2001; Gay et al., 2015). Similar phenotypic changes have been observed in compound Wnt1/Wnt3a mutants, which also show impaired formation of sensory neurons and defects in several bones and cartilage derived from the NC (Ikeya et al., 1997). Ectopic Wnt1-Cre-driven expression of stabilized β-catenin (β-cat^Ex3fl/+) results in excessive differentiation of NC progenitors into NeuroD (Neurod1) and Brn3a (Pou4f1) positive sensory neurons, thus limiting NC multipotency (Lee et al., 2004). Collectively, these data confirm the requirement of Wnt/β-catenin signaling in NC maintenance and differentiation; however, they do not clarify the role of Wnts in NC induction and specification. The apparent inconsistency might be explained in two ways: (1) the signaling pathway network governing NC induction and specification differs between mouse and other species (Barriga et al., 2015); or (2) the Wnt1-Cre driver targets the already specified NC, which does not allow investigation of earlier events of NC development.

The aim of this work was to investigate the role of Wnt/β-catenin signaling in the early development of the mammalian NC. We used a Cre line derived from the NC inducer gene Tfap2a (AP2α-Cre) (Macatee et al., 2003), in order to achieve conditional deletion of Tcf7l1 in the earliest stages of NC development. Conditional inactivation of Tcf7l1 leads to transdifferentiation of ANF into NC cells, accompanied by loss of anterior neural markers. Our data suggest that Tcf7l1, acting as a Wnt/β-catenin repressor, is required to maintain anterior neural character by restricting expansion of the Wnt-induced NC cells.

Tcf7l1 is detected in the developing mouse ectoderm prior to gastrulation and later, at E7.5, it is expressed in the anterior ectoderm and mesoderm (Korinek et al., 1998; Merrill et al., 2004). We characterized the expression of Tcf7l1 during neurulation using whole-mount in situ hybridization. Tcf7l1 mRNA was detected in the anterior neural plate at the one somite pair (s) stage, being further restricted to the ANF region and along the anterior NPB with the progress of neurulation at 3-4 s stage (Fig. 1A). We next aimed to relate the expression pattern of Tcf7l1 to that of known NC-specific genes and the area of Wnt/β-catenin signaling. Tfap2a is one of the key genes involved in NC induction, while Foxd3, Sox9 and Sox10 have been linked with NC specification (Groves and LaBonne, 2014; Simoes-Costa and Bronner, 2015). During neurulation Tfap2α protein is continuously expressed along the whole NPB, as shown by whole-mount immunofluorescence (Fig. 1B). First, Sox9 and Foxd3 transcripts were detected in the neural plate as early as the 1-2 s stage (Fig. 1C, Fig. S1A), and at 3-4 s they were expressed robustly in two stripes along the anterior NPB (Fig. 1C, Fig. S1A). The onset of Sox10 mRNA was detected later (4-5 s), in a pattern overlapping with Sox9 and Foxd3 expression (Fig. S1B). To map the activity of the Wnt/β-catenin pathway, we used BAT-gal reporter mice (Maretto et al., 2003). The BAT-gal signal, visualized by immunostaining, was found to spread gradually across the prospective mesencephalon and metencephalon from 2-3 s to 4-5 s stage, but it was never detected in the ANF (Fig. 1D). In conclusion, Tcf7l1 expression in the ANF overlaps with that of Tfap2α in the anterior part of the NPB but it is reciprocal to that of Sox9, Foxd3, Sox10 and BAT-gal, which reside in the caudal areas of the neural plate.

In order to study the function of Tcf7l1 during NC induction and specification, we inactivated Tcf7l1 by crossing Tcf7l1 conditional knockout mice (EUCOMM) with AP2α-IRESCre (AP2α-Cre) transgenics (Macatee et al., 2003). Tracing of the AP2α-Cre-driven recombination using the Rosa26^EYFPfl/+ reporter strain revealed the first EYFP⁺ cells scattered throughout the embryo at E7.0 (Fig. 1E). The EYFP signal became focused along the entire NPB at the 4 s stage, with the highest levels detected rostrally in the neural- and non-neural ectoderm of the ANF at the 5 s stage. AP2α-Cre most likely targets the entire NC population because all Sox10⁺ NC cells were included in the EYFP⁺ tissue of AP2α-Cre;Rosa26^EYFPfl/+ embryos at the 10 s stage (Fig. 1E). At E10.5, the EYFP signal was present in NC derivatives (brachial arches, dorsal root ganglia and sensory neurons), the head ectoderm, lens placodes, and in the anterior telencephalon (Fig. 1E, Fig. S1C).

In AP2α-Cre;Tcf7l1^fl/fl mutants, Tcf7l1 mRNA was completely lost, whereas control embryos showed high levels of transcripts along the anterior NPB including the ANF at the 5 s stage (Fig. 1F,G). At this stage, phenotypic changes following Tcf7l1 deletion became apparent: neural folds narrowed and the neural plate expanded laterally at the prosencephalic-mesencephalic boundary (Fig. 1F,G). Later, at E10.5, we observed a variability in the penetrance of morphological changes among individual mutants. We classified these defects into two categories (Fig. 1H): (1) a ‘strong' phenotype, represented by a severe reduction of the telencephalic tissue and exencephaly; and (2) a ‘mild' phenotype, with reduced telencephalon and bilateral anopthalmia. Quantification of phenotype incidence between E10.5 and E13.5 revealed that the strong phenotype was present in 60.5% and the mild phenotype in 39.5% of the embryos analyzed (n=38) (Fig. 1I). Mutant embryos with the strong phenotype did not survive beyond E13.5, exhibiting massive exencephaly and severe defects in craniofacial structures (Fig. S1D,E). These data demonstrate that Tcf7l1 function in the ANF is crucial for the development of the mouse forebrain.

Tcf7l1 has been implicated in A-P patterning of the developing brain in both zebrafish and mouse, since loss-of-function mutations result in severe reductions of anterior neural regions, accompanied by apparent expansion of midbrain and hindbrain markers towards the anterior (Dorsky et al., 2003; Merrill et al., 2004). In contrast to published data, we did not observe any A-P shift in mRNA expression of the core regulators of the midbrain-hindbrain boundary (MHB) Fgf8, Gbx2 and Wnt1 between Tcf7l1 conditional mutants and controls at E9.0 (Fig. 2A-B′, Fig. S3B). However, we did observe a progressive loss of the posterior prosencephalon marker Tcf7l2 mRNA (Fig. 2C-C″), illustrating a gradual loss of anterior neural tissue in AP2α-Cre;Tcf7l1^fl/fl mutants with mild (Fig. 2C′) or strong (Fig. 2C″) phenotype at E9.0. Furthermore, expression of the ANF-specific transcripts Six3 (Lagutin et al., 2003), Foxg1 (Hanashima et al., 2004) and Fgf8 (Meyers et al., 1998) was lost or strongly reduced in the mutants at the 8 s, 11 s and 17 s stage (E9.5), respectively (Fig. 2D-F′). Next, we tested whether the ANF cells in Tcf7l1 conditional mutants retained their neural character. RNA in situ hybridization revealed reduced expression of the general neural marker Sox2, and expansion of Sox2⁻ cells along the anterior lateral border of the ANF in AP2α-Cre;Tcf7l1^fl/fl mutants (Fig. 2G,G′), suggesting that these cells have lost their neural identity.

NC cells need to lose their neuroepithelial character in order to delaminate and migrate towards their target destinations. NC specification, which precedes delamination in Xenopus and chicken, is regulated by the transcription factors FoxD3, Sox9 and Sox10 (Cheung et al., 2005; Dottori et al., 2001; Liu et al., 2013). Therefore, we investigated the expression of these three factors in the absence of Tcf7l1.

As can be seen from whole-mount immunostaining, Sox9⁺ NC cells expanded into the Sox1⁺ ANF region in AP2α-Cre;Tcf7l1^fl/fl mutants compared with controls at the 5 s stage (Fig. 3A). Examination under higher magnification showed decreased neural Sox1 staining towards the anterior part of the mutant embryo. A frontal view of the same embryos clearly showed ectopic Sox9⁺ NC cells along the ANF in mutants (Fig. 3B). The loss of the neuroectodermal marker Sox1 was not caused by any increase in apoptosis, as we did not detect any changes in cleaved caspase 3 immunostaining in the mutants at 5 s stage (n=4, Fig. S2A). Upregulation of Sox9 expression was further confirmed by in situ hybridization of Sox9 mRNA (Fig. S2B). Interestingly, we also observed a cell fate shift characterized by the change from Sox1 to Sox9 in the caudal neural tissue. Coronal sectioning of the developing hindbrain at E9.0 also uncovered aberrant Sox9⁺/Sox1⁻ cells in the mutants (Fig. S2C). Consistent with the ectopic expansion of Sox9-expressing cells, the expression of Foxd3 was also upregulated along the anterior NPB in Tcf7l1 conditional mutants at the 10 s stage (Fig. 3C). Later, at E9.0, aberrant Foxd3 mRNA was detected in the rostral part of the head in mutants (Fig. S2D).

Sox10 mRNA expression was delayed in the absence of Tcf7l1, as it was barely detectable at the 6 s stage, but it emerged in elevated levels one day later at E9.5 in ectopic patches in the anterior part of the malformed head of mutant embryos (Fig. 3D). The onset of Sox10 expression in Xenopus and chicken follows after Sox9 and Foxd3, which appear immediately after NC induction, suggesting its requirement in later events of NC specification (Cheung and Briscoe, 2003; Cheung et al., 2005; McKeown et al., 2005; Sakai et al., 2006). The delay in the onset of Sox10 expression in the AP2α-Cre;Tcf7l1^fl/fl mutants could be caused by a prolonged NC induction. To verify this, we analyzed the expression of the key NC inducer Tfap2α (de Croze et al., 2011). Indeed, the region of Tfap2α immunofluorescence was robustly enlarged along the neuroectoderm of the anterior NPB in the AP2α-Cre;Tcf7l1^fl/fl mutants at the 8 s stage (Fig. 3E). Double staining of Tfap2α together with Sox10 further revealed that only a few Sox10⁺ cells managed to migrate from the NPB in the mutants (Fig. 3E), while the rest of the Sox10 signal appeared to be amassed in the Tfap2α⁺ neuroectoderm of the Tcf7l1 conditional mutants.

Finally, Pax3 mRNA signal spread aberrantly along the ANF in a similar way to Tfap2α, Sox9 and Foxd3 (Fig. 3F). We wanted to confirm this phenomenon in other vertebrates. We co-injected translation-blocking morpholinos (MOs) against zebrafish tcf7l1a and tcf7l1b into fertilized zebrafish eggs and stained for foxd3, pax3 and sox9 mRNA by whole-mount in situ hybridization 12 h post-fertilization (hpf). We observed ectopic expression of foxd3, pax3 and sox9 mRNA along the entire anterior NPB of the tcf7l1a/b morphant zebrafish (Fig. S2E). The ectopic NC induction was almost completely rescued by co-injection of mouse Tcf7l1 mRNA (Fig. S2F).

Taken together, the expression of genes responsible for NC induction and specification was expanded along the ANF in AP2α-Cre;Tcf7l1^fl/fl mouse mutants and in tcf7l1 zebrafish morphants. This suggests that Tcf7l1, by forming a boundary to the expression of NC specifiers, maintains the neuroepithelial character of the lateral ANF and restricts NC induction from the anterior head.

Tcf7l1 is a very efficient repressor of Wnt/β-catenin signaling target genes, as has been shown in both mouse embryonic stem cells and in vivo (Cole et al., 2008; Shy et al., 2013; Wu et al., 2012). Gain- and loss-of-function experiments in non-mammalian vertebrates have shown that the Wnt/β-catenin pathway is required for physiological (and sufficient for ectopic) induction and expansion of the NC (Garcia-Castro et al., 2002; Heeg-Truesdell and LaBonne, 2006; Steventon et al., 2009; Taneyhill and Bronner-Fraser, 2005; Wu et al., 2005). Tcf7l1 was able to significantly reduce the levels of Wnt reporter signal by competing with Tcf7 and Lef1 or by decreasing the pool of free ΔNβ-catenin required for transactivation by the Tcf/Lefs in the transient SuperTopFlash (STF) reporter assay in HEK293 cells (Fig. S3A). We therefore asked whether the expansion of NC cells in AP2α-Cre;Tcf7l1^fl/fl mutants was caused by upregulation of Wnt/β-catenin signaling.

The expression of Sp5 mRNA, a known Wnt/β-catenin target gene (Fujimura et al., 2007), was upregulated and expanded rostrally in the mutant embryos at the 6 s stage (Fig. 4A). Apart from its role in MHB formation (McMahon and Bradley, 1990; Thomas and Capecchi, 1990), Wnt1 is required during NC development (Ikeya et al., 1997; Saint-Jeannet et al., 1997). We found aberrant Wnt1 mRNA expression in the ANF of AP2α-Cre;Tcf7l1^fl/fl mutants at the 5-6 s stage (Fig. 4B). The ectopic pattern of Wnt1 mRNA expression along the NPB of the prosencephalon was even more apparent at E9.0 in Tcf7l1 conditional mutants (Fig. 4B), but we did not notice any changes in Wnt1 expression in the MHB (Fig. S3B). As a result we conclude that the aberrant Wnt1 signal in the ANF reflects the abnormal NC induction rather than altered A-P patterning.

The upregulation of the Wnt ligand (Wnt1) and Wnt target (Sp5) suggests an ectopic activation of Wnt/β-catenin signaling in the anterior neuroectoderm of AP2α-Cre;Tcf7l1^fl/fl mutants. To confirm this, BAT-gal reporter mice were crossed with Tcf7l1 conditional mutants and the resulting embryos were analyzed using anti-β-galactosidase and anti-Sox9 antibodies. As expected, the BAT-gal immunofluorescence expanded anteriorly in the AP2α-Cre;Tcf7l1^fl/fl mutants when compared with controls at the 8 s stage (Fig. 4C,D). This was accompanied by many ectopic Sox9⁺ cells, which were clearly visible from a frontal view of the 8 s stage embryos (Fig. 4D). Also, wild-type embryos showed many cells co-expressing Sox9 and BAT-gal (Fig. S3C), indicating that Sox9 may be induced by Wnt/β-catenin signaling.

Our data suggest that transdifferentiation of the ANF into NC cells in AP2α-Cre;Tcf7l1^fl/fl mutants results from aberrant activation of the Wnt/β-catenin pathway. To verify this by an alternative approach, we activated Wnt/β-catenin signaling using β-cat^Ex3fl/+ mice (Harada et al., 1999), in which a truncated, non-degradable form of β-catenin (encoded by Ctnnb1) is produced upon AP2α-Cre-mediated recombination. The AP2α-Cre;β-cat^Ex3fl/+ embryos exhibited neural tube closure defects and did not survive beyond E10.5. The crossing of the BAT-gal reporter into AP2α-Cre;β-cat^Ex3fl/+ mutants revealed strong Wnt reporter activation in most of the head ectoderm and along the anterior NPB (Fig. 4C,E). However, when compared with the BAT-gal signal in the AP2α-Cre;Tcf7l1^fl/fl mutants, it was clear that the levels were much weaker in the ANF of the AP2α-Cre;β-cat^Ex3fl/+ embryos (Fig. 4D,E). Again, Sox9 immunofluorescence reflected the BAT-gal expansion in AP2α-Cre;β-cat^Ex3fl/+ mutants (Fig. 4E), but it did not reach the most anterior tip of the neuroectoderm, unlike in Tcf7l1 mutants (Fig. 4D).

Next, we attempted to compare the expression of several targets of the Wnt/β-catenin signaling pathway and NC-related genes using qRT-PCR. We isolated mRNA from dissected heads of individual AP2α-Cre;β-cat^Ex3fl/+ (n=6) and AP2α-Cre;Tcf7l1^fl/fl embryos (n=6) at the 6-12 s stage (Fig. S3D) and corresponding controls β-cat^Ex3fl/+ and AP2α-Cre;Tcf7l1^fl/+ (n=5, Fig. S3E), respectively. Significantly higher levels of Lef1 (3.94-fold) and Axin2 (2.62-fold) mRNA were expressed in the β-cat^Ex3fl/+ heads when compared with AP2α-Cre;Tcf7l1^fl/fl heads, reflecting the overall wider activation of the BAT-gal reporter (Fig. S3D). However, there was no significant difference in Sp5 expression between these mutants, presumably as Sp5 is upregulated to a similar extent in both mutants. AP2α-Cre;β-cat^Ex3fl/+ did not display significant changes in the levels of Sox10, Foxd3 and Pax3 mRNA in comparison to AP2α-Cre;Tcf7l1^fl/fl mutants. Interestingly, Sox9 mRNA expression was decreased to 50.5% and Tfap2a to 43.5%, suggesting that the induction of NC in AP2α-Cre;β-cat^Ex3fl/+ was weaker than in AP2α-Cre;Tcf7l1^fl/fl heads at E8.5.

The qRT-PCR analysis of whole heads obviously concealed changes in the spatial distribution of mRNA expression. To assess this, we employed whole-mount RNA in situ hybridization of AP2α-Cre;β-cat^Ex3fl/+ embryos. Levels of Sox9 mRNA were increased and they expanded along the lateral border of the ANF in the β-cat^Ex3fl/+ mutants at the 9 s stage (Fig. S3F), similarly to the Sox9 immunofluorescence shown in Fig. 4E. Sox10 mRNA levels appeared weaker and their distribution was more disperse than in the controls at 10 s stage (Fig. S3G). At E9.0, Sox10 was aberrantly expressed in the head and along the dorsal part of the posterior trunk in AP2α-Cre;β-cat^Ex3fl/+ embryos (Fig. S3H). Overall expression of Foxd3 mRNA was elevated in β-cat^Ex3fl/+ mutants at the 8 s stage (Fig. S3I), and at E9.5 in the region roughly corresponding to that of developing sensory nerves (Fig. S3I′).

In addition to the effects on the NC population, the strong overproduction of stabilized β-catenin in AP2α-Cre;β-cat^Ex3fl/+ mutants led to disorganization of the neural tissue and the non-neural ectoderm along the NPB. This was clearly visible from the expanded expression of Fgf8 and Gbx2 mRNA at E9.5 (Fig. S3J-K). Of note, the characteristic pattern of Fgf8 and Gbx2 expression in the MHB was maintained in AP2α-Cre;β-cat^Ex3fl/+ mutants (Fig. S3J′,K) but the expression of both genes was expanded posteriorly along the NPB.

Next, we assayed NC delamination by immunofluorescence of neuroepithelial cells marked with N-cadherin and migrating NC cells labeled with Sox9 antibody. The delamination of mouse NC cells starts around the 4 s stage (Lee et al., 2013). Transverse sections of control and Tcf7l1 conditional mutant embryos at the 5 s stage revealed no difference in the onset of delamination of Sox9⁺ cells (Fig. 4F). One day later at E9.5, we observed only a thin layer of Sox9⁺ cells between the N-cadherin⁺ forebrain tissue and the epidermis of control embryos. The hindbrain, labeled with N-cadherin, was found to be completely devoid of Sox9⁺ cells in the controls (Fig. 4G,G′). By contrast, AP2α-Cre;Tcf7l1^fl/fl mutants clearly displayed aberrant Sox9⁺ cells in the hindbrain prior to delamination (Fig. 4H, Fig. S2C). Additionally, N-cadherin expression was weaker in the anterior forebrain (Fig. 4H′) and in the dorsal hindbrain (Fig. 4H) of the Tcf7l1 conditional mutants with the strong phenotype. The same analysis of AP2α-Cre;β-cat^Ex3fl/+ embryos revealed that almost all forebrain cells converted into Sox9⁺ NC (Fig. 4I′). Similarly to Tcf7l1 conditional mutants, we also detected aberrant Sox9-expressing cells in the hindbrain (Fig. 4I), accompanied by a massive decrease in the levels of N-cadherin (Fig. 4I,I′). Furthermore, protein levels of Lef1, a Wnt/β-catenin transcriptional activator and target gene (Galceran et al., 1999; Kratochwil et al., 2002; Liu et al., 2005), were increased in the forebrain of the Tcf7l1 mutants compared with controls (Fig. 4J,K). AP2α-Cre;β-cat^Ex3fl/+ mutants also displayed very strong upregulation of Lef1 in the mesenchyme adjacent to the hindbrain as well as in the forebrain (Fig. 4J,L). The whole-mount immunofluorescence of wild-type embryos revealed colocalization of Lef1⁺ and Sox10⁺ cells at the 8 s stage (Fig. S3L).

Taken together, deletion of Tcf7l1 resulted in ectopic activation of the Wnt/β-catenin pathway, suppression of neuroepithelial character and initiation of aberrant NC cell fate. Similar observations were made in conditional AP2α-Cre;β-cat^Ex3fl/+ mutants, although the onset of NC induction was delayed and less pronounced. Thus, the timing, spatial distribution and character of Wnt/β-catenin pathway activation differ slightly in the AP2α-Cre;β-cat^Ex3fl/+ and Tcf7l1 conditional mutants.

Our data suggest that the loss of Tcf7l1-mediated repression of Wnt targets is the primary cause of NC cell fate expansion. In order to verify this hypothesis, we performed rescue experiments using two mouse strains conditionally expressing either dominant Tcf repressor or activator. As a dominant repressor we used the mouse strain Rosa26^dnTCF4fl/+, which contains the N-terminally truncated form of human TCF7L2 (dnTCF4) targeted into the ROSA26 locus and inserted 3′ to the lox-STOP TdTomato (TdT) cassette (originally referred to as Rosa26^+/tdTomato) (Janeckova et al., 2016). This form cannot bind β-catenin and thus acts as a dominant repressor of Wnt/β-catenin targets. Second, we generated the mouse strain Rosa26^{ctβcat-Tcf7l1fl/+}, which allows conditional expression of full-length mouse Tcf7l1 fused with the transactivation domain of β-catenin (amino acids 696-781), analogous to the described Cat-Lef fusion (Hsu et al., 1998). This construct, which mimics β-catenin binding to Tcf7l1, was able to deliver mild activation of the Wnt reporter STF (Fig. S4A). Details of strain generation and characterization are presented in Fig. S4B-I. Following Cre recombination, the TdT cassette is excised, allowing production of either dnTCF4 or ctβcat-Tcf7l1 mRNA (Fig. 5A).

We crossed the Tcf7l1^fl/fl;Rosa26^dnTCF4fl/fl and AP2α-Cre;Tcf7l1^fl/+ strains and evaluated the strength of the phenotype in AP2α-Cre;Tcf7l1^fl/fl;Rosa26^dnTCF4fl/+ compound mutants at E10.5. Overexpression of dnTCF4 was able to reduce the occurrence of the strong phenotype by more than half (to 21%), compared with the 60% incidence in AP2α-Cre;Tcf7l1^fl/fl littermates (P=0.049, Fisher's exact test). The effect of the rescue on the NC population was further analyzed by in situ hybridization, showing no or very little ectopic Sox10 mRNA in the AP2α-Cre;Tcf7l1^fl/fl;Rosa26^dnTCF4fl/+ ‘rescued’ mutants (Fig. 5B,D). By contrast, the βcat-Tcf7l1 fusion protein showed a significantly (P=0.0016, Fisher's exact test) decreased capability to rescue the phenotype compared with dnTCF4, as demonstrated by 79% of analyzed embryos displaying the strong phenotype in AP2α-Cre;Tcf7l1^fl/fl;Rosa26^{ctβcat-Tcf7l1fl/+} compound mutants. Expression of ctβcat-Tcf7l1 was also unable to reduce the aberrant Sox10⁺ NC population (Fig. 5C,D). In summary, using transgenic mouse strains, we genetically confirmed that Tcf7l1, acting as a repressor of Wnt/β-catenin signaling, protects the anterior neural tissue from conversion to the NC cell fate.

Spatiotemporal regulation of Wnt/β-catenin pathway activation, which is required for proper development of the prosencephalon, is secured by the expression of several ANF-specific negative regulators of Wnt/β-catenin signaling. Our data, together with evidence from others (Cole et al., 2008; Liu et al., 2005; Ombrato et al., 2012; Wu et al., 2012), characterize Tcf7l1 as a repressor of Wnt/β-catenin target genes. Aberrant activation of the Wnt/β-catenin pathway in Tcf7l1 systemic mutants is regarded as the cause of strong A-P patterning defects both in zebrafish (Dorsky et al., 2003; Kim et al., 2000) and mouse (Merrill et al., 2004). tcf7l1a loss-of-function (hdl) zebrafish mutants or tcf7l1a and tcf7l1b morphants display a shift in the expression of the MHB markers pax2.1 and gbx1 towards the anterior, which is accompanied by the loss of six3 and rx3 expression (Dorsky et al., 2003). The Tcf7l1 mouse systemic mutants also display expansion of the midbrain marker En1 and downregulation of Hesx1 (Merrill et al., 2004). Similarly, deletion of Dkk1 (Mukhopadhyay et al., 2001), Six3 (Lagutin et al., 2003), Hesx1 (Andoniadou et al., 2007) and ICAT (Ctnnbip1) (Satoh et al., 2004) in mouse is followed by the loss of anterior neural tissue. This has usually been interpreted as a defect in A-P patterning induced by aberrant Wnt/β-catenin signaling. We think that this is a simplified view and that, in the context of the ANF phenotypes, it is only partially true. In Xenopus, low levels of ectopic Wnt/β-catenin activation enrich the NC population without affecting A-P patterning, whereas high doses affect both (Wu et al., 2005). This scenario might be also valid for mammals because mice carrying ENU-induced activating point mutations in Lrp6 and β-catenin display a gradual loss of the prospective prosencephalon depending on the strength of the Wnt/β-catenin pathway overactivation, but they do not manifest an anterior shift of posterior neural markers (Fossat et al., 2011). Furthermore, compound deletion of one Hesx1 allele and both alleles of Tcf7l1 in mouse, using prosencephalon-specific Hesx1-Cre, results in the aberrant activation of Wnt/β-catenin signaling that leads to a loss of anterior neural tissue, whereas A-P patterning remains unchanged (Andoniadou et al., 2011). Our observations in AP2α-Cre;Tcf7l1^fl/fl mutants are in agreement with these data as we did not detect any alteration in expression of Fgf8 or Gbx2 in the MHB (Fig. 2). Nevertheless, we did observe a gradual loss of the ANF markers Six3, Fgf8 and Foxg1 and the posterior prosencephalon marker Tcf7l2 depending on the strength of the phenotype. This was accompanied by enrichment of the Tfap2a⁺, Pax3⁺, Sox9⁺, Foxd3⁺ and Sox10⁺ NC population and its expansion along the lateral border of the anterior ANF (Fig. 3). We therefore provide a new explanation for the loss of the ANF. The absence of Tcf7l1-mediated repression of Wnt target genes in AP2α-Cre;Tcf7l1^fl/fl mutants allows the gradient of Wnt ligands to influence cell fate in a wider region of the ANF than normally. This permits NC induction in the ANF, while A-P patterning of the neural tissue remains unaltered. A different experimental design, based on Wnt/β-catenin pathway overactivation in AP2α-Cre;β-cat^Ex3fl/+ mutants, also resulted in aberrant NC induction that was spread along the A-P axis. In contrast to AP2α-Cre;Tcf7l1^fl/fl mutants, the most anterior neural fold did not convert to NC in AP2α-Cre;β-cat^Ex3fl/+ mutants (Fig. 4, Fig. S3). This might be explained by a lower activation of the Wnt/β-catenin pathway in the ANF of AP2α-Cre;β-cat^Ex3fl/+ than of AP2α-Cre;Tcf7l1^fl/fl mutants, as judged by the BAT-gal reporter (Fig. 4). We think that the presence of several factors, including Tcf7l1, antagonizing Wnt-induced transcription in the ANF may attenuate the effect of stabilized β-catenin.

The timing of Wnt/β-catenin activation in the head neuroectoderm is crucial. It has been shown in Xenopus that NPB loses its competence for NC induction by Wnts with time due to increasing BMP activity (Steventon et al., 2009; Steventon and Mayor, 2012). The use of AP2α-Cre (onset of Cre recombination prior to neurulation, later expressed along the NPB) instead of the widely used Wnt1-Cre (onset of Cre recombination around the 5 s stage, expressed in already specified NC) (Danielian et al., 1998; Lewis et al., 2013) can yield dramatically different NC phenotypes. In AP2α-Cre;β-cat^Ex3fl/+ mutants, we affected the population of undifferentiated Sox9⁺ and Foxd3-expressing NC cells as early as at the 8 s stage, whereas in Wnt1-Cre/β-cat^Ex3fl/+ mutants altered NC differentiation was observed much later at E10.5 (Lee et al., 2004). This indicates the AP2α-Cre driver to be a more suitable tool for studying early events of NC development. Our data provide evidence for an earlier role of the Wnt/β-catenin pathway in NC development than evident from previously published findings in the mouse model showing the requirement for the Wnt/β-catenin pathway in NC maintenance and differentiation (Brault et al., 2001; Gay et al., 2015; Lee et al., 2004).

The effects of the Tcf7l1 deficiency in mice and zebrafish on NC development were not, until now, characterized in detail. Analogous to the NC expansion in AP2α-Cre;Tcf7l1^fl/fl mice (Figs 3,4, Fig. S2), we observed ectopic expression of foxd3, pax3 and sox9 mRNA along the entire anterior NPB of tcf7l1a/b morphants in zebrafish. Our rescue experiments with co-injection of mouse Tcf7l1 mRNA implicate the conservation of Tcf7l1 function across vertebrate species. In accordance, MO-mediated knockdown of Dkk1 in Xenopus is followed by enrichment of the NC population, which is marked by the expansion of Snail2 towards the anterior portion of the embryo (Carmona-Fontaine et al., 2007). In mouse, loss of Dkk1 has been shown to cause expansion of the NC marker Sox10 anteriorly (Carmona-Fontaine et al., 2007). Similarly, Hesx1 mutant mice display expansion of Wnt target genes together with the NC markers Pax3 and Foxd3 towards the anterior (Andoniadou et al., 2007). Mutual functional redundancy of these factors was demonstrated in zebrafish in which injection of either Six3 or Hesx1 mRNA rescued the hdl phenotype (tcf7l1a deletion) (Andoniadou et al., 2011; Lagutin et al., 2003). All these data are in agreement with the interpretation of Tcf7l1 mutants presented here.

In summary, collective evidence points towards the coordinated expression of several Wnt/β-catenin pathway inhibitors in the anterior neuroectoderm that together define, under physiological conditions, a spatial border to the expansion of the NC. Similarity in the phenotypes following their deficiency suggests that the ANF region is very close to the threshold of Wnt/β-catenin activation for NC induction. Therefore, from the evolutionary perspective, attenuation of the sensitivity to Wnt/β-catenin signaling in the ANF region by fine-tuning the expression levels of Wnt signaling inhibitors might represent a very sensitive tool that allows the simultaneous modification of two prominent vertebrate features, namely the NC and the forebrain.

Our analysis of AP2α-Cre;Tcf7l1^fl/fl mutants identified Tcf7l1 as a negative regulator of the NC population. Given that the NC is positively regulated by the Wnt/β-catenin pathway in all vertebrates studied, other Tcf/Lefs should be responsible for activation of Wnt/β-catenin targets in the NC. It is likely that Lef1, Tcf7 and Tcf7l2 are redundant in NC cells, since whole-body double knockouts of Lef1/Tcf7 and Tcf7/Tcf7l2 in mice do not display any obvious NC phenotype (Galceran et al., 1999; Gregorieff et al., 2004). Ectopic Sox9⁺ cells in AP2α-Cre;β-cat^Ex3fl/+ mutants spatially overlap with the elevation of the BAT-gal reporter, Lef1 and Tcf7, but not Tcf7l2 proteins (Fig. 4; data not shown), suggesting that Lef1 and Tcf7 are positively regulated by Wnt/β-catenin signaling and probably participate in the expression of Wnt-responsive NC-specific genes. In Xenopus, Tfap2α is both required and sufficient for induction of NC. Additionally, expression of Tfap2a was inhibited by injection of an MO against β-catenin at stage 17 (de Croze et al., 2011), raising the possibility that Tfap2a is a Wnt/β-catenin target gene. This is in agreement with the data presented here showing upregulation of Tfap2α expression in AP2α-Cre;Tcf7l1^fl/fl mutants (Fig. 3). Sp5, another well-defined Wnt target gene and transcriptional repressor (Fujimura et al., 2007), was strongly upregulated in AP2α-Cre;Tcf7l1^fl/fl mutants, spreading its expression across the entire ANF (Fig. 4). The work in Xenopus has identified Sp5 as a novel NC-specific inducer (Park et al., 2013). MO-mediated knockdown of Sp5 leads to the upregulation of Zic1, making it a candidate target of Sp5 repression. Zic1 is able to initiate NC formation only when expressed in appropriate balance with Pax3, another NC inducer (Hong and Saint-Jeannet, 2007; Park et al., 2013). Pax3, which is also upregulated in AP2α-Cre;Tcf7l1 mutants (Fig. 3), has recently been identified as a direct target of Tcf/Lefs (Zhao et al., 2014), providing an additional link to the gene regulatory network driving NC induction. Aberrant induction of Sp5-mediated repression of Zic1, in combination with the ectopic expression of Wnt1 ligand (Fig. 4), which enhances the β-catenin/Tcf/Lef-driven transcription of Pax3, might serve as an artificial positive-feedback loop, further potentiating the effects on NC induction. Additionally, the aberrant Wnt/β-catenin signaling present in AP2α-Cre;Tcf7l1^fl/fl mutants might upregulate the expression of Axud1 (Csrnp1), a novel Wnt-dependent positive regulator of NC specification (Simões-Costa et al., 2015).

In summary, we have demonstrated that Tcf7l1-mediated repression of β-catenin target genes protects the anterior neuroectoderm from converting into NC. Thereby, we (1) help to eliminate the ambiguity in the interpretation of NC development in mouse and other vertebrate model organisms and (2) provide a model explaining the phenotypes previously acquired in mouse following derepression of the Wnt/β-catenin signaling pathway during early brain development.

Housing of mice and in vivo experiments were performed in compliance with the European Communities Council Directive of 24 November 1986 (86/609/EEC) and national and institutional guidelines. All procedures involving experimental animals were approved by the Institutional Committee for Animal Care and Use (permission PP-071/2011, PP-281/2011).

AP2α-IRESCre (AP2α-Cre) was described previously (Macatee et al., 2003). Cre reporter lines Rosa26^EYFPfl/+ [B6.129X1-Gt(ROSA)26Sor^tm1(EYFP)Cos/J, stock no. 00614] and ROSA26 [B6.129S4-Gt(ROSA)26Sor^tm1Sor/J, stock no. 003474] were purchased from The Jackson Laboratory. Tcf7l1^fl/+ strain [Tcf7l1^{tm1a(EUCOMM)Wtsi}] was purchased from the EUCOMM. Transgenic mouse strain BAT-gal (Maretto et al., 2003) was kindly provided by S. Piccolo. Mice with a conditional ‘floxed' β-catenin (floxed Ctnnb1 exon 3, referred to as β-cat^Ex3fl/+) were provided by M.M.T. (Harada et al., 1999). Generation of Rosa26^dnTCF4fl/+ was described by Janeckova et al. (2016). Briefly, human TCF7L2 coding sequence lacking the N-terminal domain (amino acids 1-31) was cloned into the ROSA26 targeting vector, which contains arms for homologous recombination into the ROSA26 locus, in the same way as when generating Rosa26^{ctβcat-Tcf7l1fl/+}. Details of Rosa26^{ctβcat-Tcf7l1fl/+} mouse line generation are provided in Fig. S4. A C-terminal part of β-catenin (amino acids 696-781) was fused with mouse full-length Tcf7l1 and inserted into the ROSA26 locus by homologous recombination in R1 embryonic stem cells (Stem Cell Technologies). The construct carries tdTomato (kindly provided by Roger Tsian, UC San Diego, CA, USA) coding sequence flanked by LoxP sites serving as a stop cassette for ctβcat-Tcf7l1. Excision of the tdTomato cassette via Cre-mediated recombination allows transcription of ctβcat-Tcf7l1 mRNA (Fig. 5A). The PGK-Neo cassette was removed by crossing F1 animals with the Actb:FLPe mouse strain expressing FLP-FRT recombinase (Rodríguez et al., 2000), backcrossed with C57BL/6J for six generations. All above-mentioned mouse strains were kept on a C57BL/6J background.

Primers (5′-3′) for genotyping the Rosa26^{ctβcat-Tcf7l1fl/+} allele: JM21F, GAGGAGCCATAACTGCAGAC; JM23R, CCTGGCTATCCTAGATAGAAC; and JM27R, GTACTGTTCCACGATGGT. The expected size of the PCR product is 965 bp in the mutant and 300 bp in the wild type. Recombination in AP2α-Cre;Tcf7l1^fl/fl;Rosa26^{ctβcat-Tcf7l1fl/+} was detected using a mix of JM21F, JM27R and JM05B GATCAGCTCGTCGTTCGCTC primers, resulting in PCR products of 970 bp prior to, and 777 bp after, recombination.

Zebrafish wild-type, single-cell embryos were injected with a combination of previously described tcf7l1a (5′-CTCCGTTTAACTGAGGCATGTTGGC-3′) and tcf7l1b (5′-CGCCTCCGTTAAGCTGCGGCATGTT-3′) MOs (Dorsky et al., 2003) (Gene Tools) in a 1:1 ratio at a final concentration of 1.6 ng/nl (1-2 nl injected). Mouse Tcf7l1 mRNA used in rescue experiments was synthesized using the mMESSAGE mMACHINE Kit (Ambion) and 10 pg was injected. Embryos were raised at 28°C and harvested at 12 hpf.

At least three embryos of each genotype and similar developmental stage were analyzed per staining. Primary antibodies are listed in Table S1. Secondary antibodies were anti-mouse, anti-chicken, anti-rabbit or anti-goat Alexa Fluor 488 or 594 (1:500; Life Technologies) and biotinylated anti-rabbit (1:500; Vector Laboratories), with the Vectastain ABC Elite Kit and ImPACT DAB substrate (all Vector Laboratories). For whole-mount immunohistochemistry we used the protocol of Sandell et al. (2014). Briefly, embryos were fixed overnight in 4% paraformaldehyde (PFA), permeabilized using Dent's bleach and 100% methanol and rehydrated. Blocking for 2 h and antibody incubation overnight were performed in 3% BSA in 0.1 M Tris-Cl pH 7.5, 0.15 M NaCl, 0.03% Triton X-100. Washing five times for 1 h each in PBS was followed by overnight incubation with secondary antibody in the blocking buffer. The next day, embryos were stained for nuclei using Hoechst (1:1000; Life Technologies), washed three times in PBS (20 min each) and stored in 2% formaldehyde (Sigma). Whole-mount immunofluorescence was captured using an Apotome.2 Ax10 zoom.V16 (Zeiss) binocular microscope. Fluorescence and brightfield light images were acquired with a Nikon Diaphot 300 with 4×/0.1 NA and 10×/0.25 NA objectives and a Leica SP5 confocal microscope.

Whole-mount in situ hybridization was performed using standard protocols. Riboprobes were cloned into pGEM-T-Easy vector (Promega) using the primers specified in Table S1.

HEK293 cells (30,000 cells/well of a 96-well plate) were transiently transfected using Lipofectamine 2000 (Invitrogen) with 30 ng Super8X TopFlash (kind gift from Randall Moon) and 1 ng Renilla luciferase constructs and the indicated amount of expression vector, supplemented with carrier DNA to a total final concentration of 200 ng DNA per well. Luciferase assays were performed 48 h later with a Dual-Glo Luciferase Kit (Promega). The luciferase signal was normalized to the co-transfected Renilla readout. Experiments were repeated at least three times, using biological replicates.

To analyze mRNA expression in AP2α-Cre;β-cat^Ex3fl/+ and AP2α-Cre;Tcf7l1^fl/fl heads, eyes from 6-12 s stage (E8.5) mouse embryos were dissected and total RNA was isolated with Trizol Reagent (Life Technologies). Random-primed cDNA was generated from 200 ng total RNA using the SuperScript VILO cDNA Synthesis Kit (Life Technologies). Six different embryos originating from three litters were used for tissue dissection and subsequent analysis per genotype. qRT-PCR reactions were run in a LightCycler 480 instrument (Roche) using a 5 μl reaction mixture of DNA and SYBR Green I Master (Roche). Analysis was performed on replicates of six individual samples per genotype, run in duplicate. The average Cp values of all biological and technical replicates were normalized to endogenous levels of Tbp. Statistical significance of the change in mRNA expression was calculated by a two-tailed Student's t-test. Change in mRNA expression was presented as the ratio of AP2α-Cre;β-cat^Ex3fl/+ to AP2α-Cre;Tcf7l1^fl/fl on a log₂ scale. Primer sequences are listed in Table S1.

We thank K. Kovacova, J. Lachova, L. Svobodova and V. Noskova for excellent technical support; F. Dupacova for help with paraffin sectioning; and S. Piccolo for providing BAT-gal.