HESX1- and TCF3-mediated repression of Wnt/β-catenin targets is required for normal development of the anterior forebrain

Suppression of posteriorising signals, and in particular of Wnt signalling, is necessary for correct forebrain specification (Buchert et al., 2010; Felix and Aboobaker, 2010; Fredieu et al., 1997; Heisenberg et al., 2001; Houart et al., 2002; Kimura et al., 2000; Kudoh et al., 2002; Perea-Gomez et al., 2001; van de Water et al., 2001; Wilson and Houart, 2004). Inhibitors that are able to sequester Wnt molecules or to bind their receptors are secreted either locally within the neuroectoderm (Houart et al., 2002) or from underlying tissues (Glinka et al., 1997; Kazanskaya et al., 2000; Piccolo et al., 1999) to abolish Wnt signalling in the forebrain. In addition, other molecules act within the receiving cell to either bind to β-catenin (Garaventa et al., 1999; Satoh et al., 2004) or to interfere with receptor maturation (Yamamoto et al., 2005) to ensure that the pathway is inactive within the anterior neural plate. The combination of these cell- and non-cell-autonomous mechanisms is thought to lead to the establishment of a high-posterior to low-anterior gradient of Wnt signalling that provides the positional information required for the regionalisation of the incipient neural plate into the fore, mid and hindbrain. However, less is known about how Wnt/β-catenin target genes are repressed in forebrain precursors.

β-catenin, the product of Ctnnb1, is at the centre of the canonical Wnt pathway. In the absence of Wnt molecules, β-catenin is phosphorylated by a ‘destruction complex’ formed by several proteins, including GSK3β, CK1α, AXIN1 and APC, and targeted for ubiquitylation and subsequent proteasome-mediated degradation. Binding of secreted Wnt ligands to Frizzled and LRP membrane receptors causes the disassembly of the destruction complex and inhibition of β-catenin phosphorylation. This results in the cytoplasmic accumulation of β-catenin, which can enter the nucleus to interact with DNA-binding TCF/LEF transcription factors and activate transcription of Wnt target genes (reviewed by van Amerongen and Nusse, 2009).

Among the four TCF/LEF factor family members, only Tcf3 (Tcf7l1) is expressed in the developing forebrain primordium of the mouse at presomitic and early somite stages (Galceran et al., 1999; Merrill et al., 2004). Experiments in Xenopus laevis indicate that tcf3 acts as a repressor of Wnt targets and interacts with Groucho co-repressors (Brannon et al., 1999; Brantjes et al., 2001; Houston et al., 2002). In zebrafish, knockdown of both paralogues of Tcf3, tcf3a (tcf7l1a) and tcf3b (tcf7l1b), results in loss of repressor activity and in anterior truncations (Kim et al., 2000), indicating that there is a requirement for the β-catenin-independent Tcf repressor activity in head morphogenesis (Dorsky et al., 2003). In mouse, Tcf3-deficient embryos undergo gastrulation but exhibit variable degrees of defects, including primitive streak and axis duplications, supernumerary neural folds, and neural patterning defects involving expansion of midbrain at the expense of forebrain and hindbrain tissues (Merrill et al., 2004). However, the mechanisms underlying these defects are not fully understood because Tcf3 is expressed prior to the onset of gastrulation in the epiblast, anterior mesendoderm and anterior neuroectoderm, and defects in any of these could lead to aberrant neural patterning. Specifically, whether Tcf3 plays a role within forebrain tissue remains unknown (Merrill et al., 2004).

HESX1 is a paired-like homeobox transcription factor that is expressed in the anterior regions of the vertebrate embryo but is absent from invertebrates, including close relatives such as ascidians, amphioxus and Ciona intestinalis (Kazanskaya et al., 1997; Martinez-Barbera et al., 2000). Using genetic fate mapping, we have previously shown that derivatives of Hesx1-expressing precursors that are normally destined to populate the anterior forebrain (cerebral cortex, basal ganglia, ventral diencephalon and eyes) change their fate in Hesx1-deficient embryos and colonise posterior forebrain regions as well as the neural crest lineage (Andoniadou et al., 2007). A similar posterior transformation of anterior forebrain is observed in knockdown experiments of the Hesx1 orthologue Xanf in Xenopus (Ermakova et al., 1999). The molecular mechanisms responsible for the lack of anterior identity and cell fate transformation are not known. Understanding the pathogenesis of these early defects is also of clinical relevance, as mutations in HESX1 result in forebrain, eye and pituitary defects in humans (Dattani et al., 1998; Sajedi et al., 2008).

In this study we sought to specifically address the molecular function of Hesx1 in anterior forebrain precursors. Combining genetic and molecular approaches we reveal a novel role for Hesx1 as an antagonist of the Wnt/β-catenin pathway in the mouse and zebrafish forebrain. In addition, we demonstrate a requirement for Tcf3 in forebrain progenitors, where it genetically interacts with Hesx1 to promote anterior character by repressing the transcriptional activation of Wnt/β-catenin target genes.

Wild-type and zhdl^m881 zebrafish embryos were raised at 28°C and staged according to Kimmel et al. (Kimmel et al., 1995). Single-cell embryos were injected with 5 nl of 0.5 pmol/nl hesx1 morpholino (5′-TGCAAGAGAAGCCATTGCTAAACTC-3′, GeneTools) and/or 1 pg/nl mouse Hesx1 mRNA. hdl^m881 mutant embryos were genotyped as described (Kim et al., 2000).

Hesx1-Cre, Ctnnb1-lox(ex2-6), Tcf3-flox, R26-YFP, BAT-gal, Six3-lacZ, Hesx1^+/– and Ctnnb1^+/– mice have been described previously (Andoniadou et al., 2007; Brault et al., 2001; Lagutin et al., 2003; Maretto et al., 2003; Nguyen et al., 2009; Srinivas et al., 2001; Dattani et al., 1998; Haegel et al., 1995). Tcf3^fl/+ animals were crossed with the Actb-Cre strain to generate Tcf3^+/– animals, which were bred further on C57BL/6 in order to remove the Cre transgene from the background. Breeding of genetically modified animals and all animal procedures were carried out under the UK Home Office Animals (Scientific Procedures) Act 1986. A mixed background, backcrossed onto C57BL/6, was used for all strains. Embryos and pups were genotyped by PCR on DNA from yolk sacs, tail buds or ear biopsies as described previously (Andoniadou et al., 2007). Briefly, the thermal profile comprised a single step for 2 minutes at 94°C, followed by 35 cycles of 94°C for 30 seconds, 60°C for 30 seconds and 72°C for 45 seconds. The wild-type and mutant alleles yield bands of ∼500 bp and 250 bp, respectively. For primers, see supplementary material Table S9.

The Hesx1-eGFP targeting vector was generated using homologous regions obtained from plasmids carrying the mouse Hesx1 gene (Dattani et al., 1998). A cassette containing eGFP followed by (1) four SV40 and one PGK polyadenylation sites flanked by loxP sequences, (2) the diphtheria toxin A (DTA) coding sequence (Ivanova et al., 2005) and (3) a PGK-Neo cassette flanked by frt sequences (Andoniadou et al., 2007) was cloned into a vector containing ∼6.5 kb and 1.3 kb of 5′ and 3′ homologous regions, respectively (Fig. 5). The linearised targeting vector was electroporated into CCE ES cells (129/SvEv; kindly provided by E. Robertson, Sir William Dunn School of Pathology, Oxford, UK) and ∼400 colonies were picked, expanded and screened by PCR and Southern blot, as described previously (Andoniadou et al., 2007). Two correctly targeted clones were isolated and injected into blastocysts from C57BL/6J (Harlan) mice. Male chimeras were backcrossed to C57BL/6J females to establish the F1 generation of heterozygous mice. F1 animals were crossed to the Actb:FLPe strain (Rodriguez et al., 2000), kept on a C57BL/6J background, to excise the PGK-Neo cassette. After backcrossing with C57BL/6J animals to remove the FLPe transgene, Hesx1^eGFP/+ heterozygotes were kept on a C57BL/6J background.

For microarray gene profiling, 3- to 5-somite embryos from Hesx1^eGFP/+ intercrosses were selected for eGFP expression, phenotyped and grouped into normal (Hesx1^eGFP/+) and mutant (Hesx1^eGFP/eGFP). Pieces of each embryo were retained for genotyping to ensure the integrity of each pool. Embryos were manually dissociated, and single cells were flow sorted using a MoFlo XDP (Beckman Coulter, Fullerton, CA, USA) directly into Buffer RLT (Qiagen) for RNA extraction. Fluorescence was detected using a 530/540 filter. Cell sorting data were analysed using Summit software (Dako).

Total RNA was isolated using the RNeasy Micro Kit (Qiagen) according to the manufacturer’s recommendations with the addition to the lysis buffer of 500 ng bacterial ribosomal RNA (Roche) per reaction as carrier. cDNA synthesis, linear amplification and labelling of cRNA were carried out according to the manufacturer’s protocols using the GeneChip 3′ IVT Express Kit (Affymetrix). Gene expression profiling was performed to compare eGFP-positive purified anterior forebrain precursors from Hesx1^eGFP/+ (normal) and Hesx1^eGFP/eGFP embryos on the Affymetrix Mouse430_2 platform using the GeneChip Hybridisation, Wash and Stain Kit (Affymetrix). This analysis was carried out in triplicate for each genotype, over a ∼1 year period, using independently isolated pools of cells from several embryos (n=20-30 for each replicate) and was validated on independent biological samples by qRT-PCR (supplementary material Fig. S3). Gene expression data are deposited at ArrayExpress with ID: E-MEXP-2586. Files were processed in MATLAB (MathWorks) and GeneSpring GX (Agilent Technologies). GC-RMA normalisation was carried out and significantly differentially expressed genes were identified after the Benjamini and Hochberg false discovery rate was applied as a multiple testing correction method. GeneSpring Gene Ontology and DAVID databases were used to generate pathway lists. Where multiple probe sets were available for a single gene, the value of the unique probe set was used, or, in the case of multiple unique probe sets, that with the highest raw intensity levels.

Wholemount in situ hybridisation, X-Gal staining and immunofluorescence on paraffin sections were performed as previously described (Andoniadou et al., 2007). Antibodies against GFP (Invitrogen, 1:350), cleaved caspase 3 (Cell Signaling Technology, 1:200) and phospho-histone H3 (Upstate, 1:300) were detected with Alexa Fluor-conjugated secondary antibodies (Invitrogen, 1:350).

RNA was extracted using the RNeasy Micro Kit (Qiagen) following the manufacturer’s protocols and including on-column DNaseI digestion. Up to 2 μg RNA was used for cDNA synthesis using Omniscript reverse transcriptase (Qiagen) with random hexamers (Promega).

qRT-PCR was carried out to validate microarray results, on independent biological replicates in triplicate. Briefly, 10-50 ng template cDNA was used per reaction on an ABI 7500 Fast Cycler employing SYBR-based technology using MESA Blue reagent (Eurogentec) and specific primers (supplementary material Table S10). The thermal profile comprised a single step for 2 minutes at 50°C to activate UNG (uracil-N-glycosylase) and destroy contaminating uracil-containing template, followed by 10 minutes at 95°C to denature UNG and activate Taq polymerase, then 40 cycles of 95°C for 15 seconds and 60°C for 1 minute. Results were normalised to endogenous levels of Gapdh and were analysed using the ΔΔCt method.

We assessed excision in the forebrain by quantitative PCR on DNA using primers specifically designed against exon 3 of Ctnnb1, which should be excised in Ctnnb1-lox(ex2-6) embryos in the presence of Cre recombinase. DNA was extracted from dissected forebrains (anterior to the telencephalic/diencephalic boundary) of 9.5 dpc embryos using the DNA Micro Kit (Qiagen) using standard protocols. Primers against Gapdh provided an endogenous control and results were analysed using the ΔΔCt method.

Images of live embryos or after fixation or wholemount in situ hybridisation were captured using a Leica TCS camera and IM50 software (Leica). Images were processed using Photoshop (Adobe).

To test whether HESX1 acts to maintain anterior forebrain identity by antagonising Wnt signalling, we investigated the effect of reducing hesx1 (previously anf) levels in zygotic headless (hdl; tcf3a, tcf7l1a) zebrafish mutants. The hdl phenotype, which results from a mutation in one of the two zebrafish tcf3 paralogues, was initially characterised as a maternal zygotic (mz) mutant that displayed a severe anterior truncation phenotype, including the absence of eyes (Kim et al., 2000). However, this phenotype is not observed in zygotic (z) hdl mutants owing to compensation by tcf3b, which is also expressed in the anterior neural plate of the zebrafish embryo (Dorsky et al., 2003) (Fig. 1B). Indeed, the severe mzhdl forebrain defects can be induced in zhdl mutants upon injection of tcf3b-specific morpholinos (MOs) (Dorsky et al., 2003). We took advantage of this ‘sensitised’ zhdl background to investigate the function of hesx1 in zebrafish.

Injection of hesx1 MO into wild-type embryos had no effect (Fig. 1C; 2.5 pmol/embryo, n>100). However, when we injected embryos from hdl^+/– intercrosses, we found anterior forebrain defects in ∼25% of cases, which were confirmed to be homozygous mutants by hdl genotyping (Table 1; Fig. 1D). No defects were induced in other genotypes. The absence of eyes was apparent in hdl^–/– hesx1 morphant embryos, as was the absence of the telencephalon and part of the diencephalon, indicating that the phenotype is similar to the mzhdl phenotype (Kim et al., 2000). This phenotype was specific to the knockdown of Hesx1, as co-injection of hesx1 MO in embryos from hdl^+/– intercrosses together with mouse Hesx1 mRNA (which lacks the MO target) resulted in the complete rescue of forebrain defects of zhdl^–/– hesx1 morphants (Table 1; Fig. 1F). This confirms that the role of Hesx1 in the forebrain is conserved in both species.

In conclusion, these results reveal that loss of Hesx1 is capable of inducing a phenotype in hdl^–/– embryos that is consistent with an overactivation of Wnt signalling, making it likely to act as an antagonist of the Wnt pathway. Our results also suggest that a gene interaction between hesx1 and tcf3 is required during normal forebrain development of the zebrafish embryo.

To investigate whether a genetic interaction between Hesx1 and Tcf3 is conserved in mammals, we generated mice carrying a specific gene dosage of Hesx1 and Tcf3.

Hesx1^Cre/+ heterozygous animals carry a null allele in which the gene encoding Cre recombinase replaces the entire Hesx1 coding region. These mice are mostly normal, although there is a low (<5%) incidence of eye defects (Table 2) (Andoniadou et al., 2007). The presence of Cre is not relevant for this experiment and Hesx1^Cre/+ embryos are equivalent to Hesx1^+/–. Tcf3^+/– mice carry a null allele and are also viable and fertile (Merrill et al., 2004). By contrast, Hesx1^Cre/+;Tcf3^+/– or Hesx1^+/–;Tcf3^+/– double-heterozygous embryos from crosses between Hesx1^Cre/–;Tcf3^+/– and Hesx1^+/– showed partial penetrance and displayed a variable range of anterior defects that were classified into two categories: class I defects, where embryos showed only unilateral microphthalmia and the telencephalon developed normally; and class II defects, in embryos with bilateral microphthalmia and/or unilateral anophthalmia in conjunction with a reduction in size of the telencephalic vesicles (n=33; Table 2; supplementary material Fig. S1C-E).

In situ hybridisation for specific regional markers revealed a reduction of their expression domains concomitant with the progressive loss of anterior tissue. Expression of the forkhead box-containing transcription factor Foxg1, which is essential for normal development of the telencephalon, was markedly reduced only in class II embryos (n=5; Fig. 2E). Likewise, expression of Fgf8, a crucial signalling molecule during forebrain development (Meyers et al., 1998; Shimamura and Rubenstein, 1997), at the anterior tip of the telencephalon (ANR, anterior neural ridge) was reduced only in class II Hesx1^Cre/+;Tcf3^+/– embryos (n=6; Fig. 2D,D′). The expression domain of Pax6, a paired box transcription factor involved in forebrain patterning and boundary formation, was also reduced in the telencephalon and eye of class I and II embryos but extended normally to the posterior forebrain-midbrain boundary, suggesting that more caudal regions were unaffected (n=7; Fig. 2F).

To reduce gene dosage further, Hesx1^Cre/–;Tcf3^+/– embryos were generated. Hesx^Cre/–;Tcf3^+/+ embryos usually displayed a fully penetrant phenotype with variable expressivity, extending from class II to a group with more severe defects than those described above, termed class III, which include bilateral anophthalmia as well as a reduction in telencephalic tissue (Andoniadou et al., 2007) (Table 2; supplementary material Fig. S1F,G). When a copy of Tcf3 was removed in Hesx1^Cre/–;Tcf3^+/– embryos, the severity was dramatically increased, resulting in loss of most of the anterior forebrain, which was designated a class IV defect (n=13; Table 2; Fig. 2G-I; supplementary material Fig. S1H,I). Wholemount in situ hybridisation against Foxg1 and Pax6 demonstrated the specific loss of anterior forebrain (n=3 per marker; Fig. 2H,I). Likewise, the expression domain of Fgf8 at the ANR was absent, but normal Fgf8 staining was observed at the mid-hindbrain boundary in all embryos analysed, confirming the loss of forebrain but not midbrain tissue (n=3; Fig. 2G,G′).

Together, these experiments suggest that a minimum gene dosage of Hesx1 and Tcf3 is required for normal development of the forebrain.

In mouse embryos, Hesx1 is initially expressed in the anterior visceral endoderm at the onset of gastrulation at 6.5 dpc, in the anterior mesendoderm by 7.5 dpc and subsequently in the anterior neural ectoderm from 7.5-8.0 dpc (Thomas and Beddington, 1996). Tcf3 is expressed in the epiblast at pre-gastrulation stages and in the anterior mesendoderm and anterior neural plate at 7.5 dpc in a broader domain than Hesx1 (Merrill et al., 2004). In zebrafish, hesx1 and tcf3a show similar expression domains, comprising mainly the anterior neural plate with weak expression in the anterior mesendoderm (Dorsky et al., 2003; Kim et al., 2000; Spieler et al., 2004). Therefore, the genetic interaction between Hesx1 and Tcf3 might occur at the anterior mesendoderm and/or the anterior neural ectoderm, where both factors are expressed (Merrill et al., 2004).

We aimed to define more precisely the tissue in which this interaction is required by using two mouse strains: (1) a conditional Tcf3 mouse line in which exon 2 is flanked by loxP sites, generating a frameshift and premature termination of the protein upon Cre-mediated excision (Nguyen et al., 2009); and (2) the Hesx1-Cre knock-in strain described previously (Andoniadou et al., 2007). First, we assessed the pattern of Cre activity in Hesx1^Cre/+;R26^YFP/+ embryos at the presomitic and early somite stages. Abundant Cre-mediated excision, revealed by YFP expression, was first observed in the neuroectoderm in the 1- to 2-somite embryo, but little or no YFP expression was detected in the anterior mesendoderm-derived tissues such as the anterior foregut (supplementary material Fig. S2A; n=5 Hesx1^Cre/+;R26^YFP/+ embryos). This indicates that the Hesx1-Cre mouse strain drives Cre expression predominantly in the rostral neuroectoderm fated to become the anterior forebrain (supplementary material Fig. S1B) (Andoniadou et al., 2007). Second, we analysed whether the expression of Dkk1 and Shh, two axial mesendoderm markers required for normal brain formation, could be affected in Hesx1^Cre/+;Tcf3^fl/– embryos. In situ hybridisation revealed no significant differences in mRNA expression of these two markers in Hesx1^Cre/+;Tcf3^fl/– or Hesx1^Cre/+;Tcf3^+/– embryos compared with controls (supplementary material Fig. S2B,C; data not shown). These data suggest that the axial mesendoderm is unlikely to be affected in Hesx1^Cre/+;Tcf3^fl/– or Hesx1^Cre/+;Tcf3^+/– embryos and, together with the Cre-mediated excision analysis, support the view that conditional excision of Tcf3 through the Hesx1-Cre allele, as well as the gene interaction between these two transcription factors, are likely to occur solely in the anterior neuroectoderm.

Approximately 80% of Hesx1^Cre/+;Tcf3^fl/– embryos showed partially penetrant anterior forebrain defects ranging from normal development to unilateral microphthalmia/anophthalmia and/or small telencephalic vesicles (Table 3, class I-III; Fig. 3H-J), but ∼20% of mutants displayed forebrain truncations comparable to those observed in Hesx1^Cre/–;Tcf3^+/– embryos (Table 3, class IV; Fig. 3E-G). Proliferation and apoptosis analyses revealed comparable levels between mutants and controls (supplementary material Fig. S3), as previously shown in Hesx1^–/– embryos (Andoniadou et al., 2007). Molecular analyses using Foxg1, Fgf8 and Pax6 markers revealed a reduction in their expression domains correlating with defect severity, in keeping with the loss of forebrain tissue at 9.5 dpc (n=13; Fig. 3E-J). Despite this reduction, the zona limitans intrathalamica was detected by Shh expression and the dorsal thalamus and hypothalamus were patterned properly as assessed by Gbx2 and Nkx2-1 expression, respectively (supplementary material Fig. S4).

Since, in zebrafish, tcf3-mediated repression of Wnt/β-catenin targets is required for normal brain development (Dorsky et al., 2003; Kim et al., 2000), we reasoned that if a derepression of Wnt targets is taking place due to the loss of Tcf3 in the anterior forebrain, hence activating the pathway, this should be evidenced by anteriorisation of targets activated by Wnt expression. The direct Wnt/β-catenin/TCF/LEF target Sp5 (Weidinger et al., 2005) is normally expressed in the midbrain but is absent from the anterior forebrain (Fig. 3C). By contrast, the Sp5 expression domain was rostrally expanded in the anterior forebrain of Hesx1^Cre/+;Tcf3^fl/– embryos (n=4; Fig. 3D). Expression of Six3, an essential repressor during normal forebrain development, is negatively regulated by the Wnt signalling pathway (Braun et al., 2003; Lagutin et al., 2003). In agreement with this notion and the ectopic expression of Sp5 in the anterior forebrain, the Six3 expression domain in the anterior neural plate was markedly reduced in Hesx1^Cre/+;Tcf3^fl/– embryos at the 5- to 8-somite stage (Fig. 3B,B′). Often, Six3 expression was asymmetric in Hesx1^Cre/+;Tcf3^fl/– embryos (Fig. 3B). It is important to note that in the anterior forebrain of Hesx1^–/– mutants, Sp5 is ectopically expressed, whereas the Six3 expression domain is reduced and asymmetric defects are also observed (Andoniadou et al., 2007; Dattani et al., 1998).

In summary, three main conclusions can be drawn from these results: (1) in mouse, Tcf3 is required in anterior forebrain precursors, where it prevents the ectopic expression of Wnt/β-catenin targets; (2) there is a genetic interaction between Hesx1 and Tcf3 in the anterior neuroectoderm; and (3) the similar molecular and morphological defects observed in Hesx1^Cre/–, Hesx1^Cre/+;Tcf3^fl/– and Hesx1^Cre/+;Tcf3^+/– embryos suggest that both factors synergise in a gene dosage-dependent manner to maintain the anterior character of anterior forebrain progenitors.

The genetic analyses described above on Hesx1-deficient and Hesx1/Tcf3 compound embryos suggest that the posteriorisation of anterior forebrain precursors leading to anterior truncations is driven by the derepression of Wnt/β-catenin targets and the ectopic activation of the pathway in the anterior neural plate. To test this hypothesis further, we used the BATgal transgenic mouse line, which provides a lacZ reporter for active β-catenin/TCF/LEF signalling (Maretto et al., 2003). X-Gal staining of Hesx1^Cre/+;BATgal and BATgal embryos revealed no significant differences in the β-galactosidase staining pattern and the anterior forebrain remained unstained, demonstrating the low or absent signalling mediated by β-catenin/TCF/LEF in the anterior region of the neural plate (Fig. 4A,A′; data not shown). By contrast, Hesx1^Cre/–;BATgal mutant embryos displayed a clear anteriorisation of X-Gal staining into the rostral neural plate, relative to controls (Fig. 4B,B′). This ectopic β-galactosidase expression in the absence of Hesx1 strongly suggests that there is an activation of Wnt/β-catenin signalling in the anterior forebrain of these mutants.

We reasoned that downregulation of this signalling pathway would result in the improvement of the defective patterning and development of the anterior neural plate in Hesx1-deficient mutants. We used a genetic approach to reduce the levels of β-catenin (Ctnnb1) expression, and therefore ameliorate the levels of Wnt/β-catenin signalling in Hesx1-deficient mutants. Previously described Hesx1^+/– and Ctnnb1^+/– strains, each carrying a null allele, are both viable and fertile (Dattani et al., 1998; Haegel et al., 1995). We generated Hesx1^+/–;Ctnnb1^+/– compound mice, which were also normal and fertile. By crossing these animals to Hesx1^+/– mice we generated Hesx1^–/–;Ctnnb1^+/– and control Hesx1^–/–;Ctnnb1^+/+ embryos. Morphological comparisons did not reveal significant differences in the severity of anterior forebrain defects between these two genotypes (supplementary material Table S1; n=181 embryos), and haploinsufficiency of Ctnnb1 did not restore forebrain development in the Hesx1^–/–;Ctnnb1^+/– mutants relative to Hesx1^–/–;Ctnnb1^+/+ controls.

To remove both copies of β-catenin, we used a conditional approach resulting in the specific deletion of Ctnnb1 only in Hesx1-expressing cells. By crossing the Hesx1-Cre driver with a Ctnnb1 loss-of-function strain in which exons 2-6 are flanked by loxP sites [Ctnnb1-lox(ex2-6), hereafter Ctnnb1-LOF] (Brault et al., 2001), we overcame the gastrulation defect of Ctnnb1^–/– embryos (Haegel et al., 1995). We generated embryos completely lacking Hesx1 and conditionally null for Ctnnb1 in the forebrain (Hesx1^Cre/–;Ctnnb1^LOF/–) through crossing Hesx1^Cre/+;Ctnnb1^+/– with Hesx1^+/–;Ctnnb1^LOF/+ animals (supplementary material Fig. S5). Analysis of Hesx1^Cre/–;Ctnnb1^LOF/– embryos between 9.5 dpc and 15.5 dpc demonstrated a shift in the range of forebrain defects to a milder spectrum (ranging from normal to class II, Table 4; n=12) than those observed in Hesx1^Cre/–;Ctnnb1^LOF/+, Hesx1^Cre/–;Ctnnb1^+/+ or Hesx1^Cre/–;Ctnnb1^+/– littermates (classes II-III, n=45) (Table 4; supplementary material Table S2; Fig. 4D; P<0.001, two-tailed Fisher’s exact test).

This phenotypic rescue of forebrain development was also supported by an overall improvement of anterior neural plate patterning in Hesx1^Cre/–;Ctnnb1^LOF/– embryos. The Fgf8 expression domain in the ANR of Hesx1^Cre/– mutants is small and appears restricted to the medial region of the neural plate (Fig. 4H; n=3). In Hesx1^Cre/–;Ctnnb1^LOF/– ‘rescued’ mutants, we observed a lateral expansion of the Fgf8 expression domain (Fig. 4G); in some cases, this improvement was asymmetric, consistent with a frequent asymmetric improvement of forebrain tissue at later stages. Related to this, expression of the anterior forebrain marker Six3 occurred over a broader domain of anterior neural plate in rescued embryos as compared with Hesx1 null mutants, as more forebrain tissue was correctly specified at 8.5 dpc (Fig. 4J; n=4). Whereas in Hesx1 null mutants, expression of the Wnt target Sp5 was rostrally expanded, reaching the anterior tip of the neural plate even prior to any evidence of a reduction in tissue (Andoniadou et al., 2007) (Fig. 4O,P; n=6), in Hesx1^Cre/–;Ctnnb1^LOF/– mutants at 8.5 dpc, no Sp5 transcripts were detected in the anterior forebrain, confirming a restoration of normal patterning (Fig. 4N; n=3).

Together, these results demonstrate that ectopic anterior activation of the Wnt/β-catenin signalling pathway plays an essential role in the pathogenesis of the forebrain defects in Hesx1-deficient mutants.

Our results demonstrate that conditional removal of Ctnnb1 in the anterior forebrain leading to a reduction of Wnt/β-catenin signalling is sufficient to improve forebrain patterning in Hesx1^Cre/– mutants, strongly suggesting that the mechanisms underlying the forebrain defects in Hesx1-deficient embryos are mediated by the ectopic activation of Wnt/β-catenin signalling in the anterior neural plate. We investigated this hypothesis further by performing gene profiling analyses, comparing anterior forebrain precursors expressing and not expressing Hesx1 at the 3- to 5-somite stage, just when the first morphological and molecular abnormalities are detectable in Hesx1-deficient mutants. Because the Hesx1 expression domain is very restricted at this stage, we generated a Hesx1-eGFP knock-in mouse line by replacing the Hesx1 coding region with eGFP as a tool to flow sort this small population of forebrain progenitors (Fig. 5A). This enabled us to avoid the contamination from Hesx1 non-expressing cells that would be present if performing this analysis using whole embryos or micro-dissected regions.

Hesx1^eGFP/+ mice were normal and fertile and embryos showed fluorescence in the anterior neural plate (Fig. 5C) in a pattern identical to endogenous Hesx1 expression. By contrast, Hesx1^eGFP/eGFP embryos (Fig. 5D) showed the anterior forebrain defects observed in Hesx1-deficient embryos and the pattern of fluorescence was faithful to the residual anterior forebrain domain in these mutants, which is restricted to the most anteromedial region of the neural plate and is marked by Six3 expression (Martinez-Barbera et al., 2000).

Microarray analysis of anterior forebrain precursors isolated by flow sorting from Hesx1^eGFP/+ and Hesx1^eGFP/eGFP 3- to 5-somite embryos (Fig. 5) confirmed changes in gene expression that we had previously characterised by wholemount in situ hybridisation on Hesx1 null mutants (Andoniadou et al., 2007). For example, by wholemount in situ hybridisation, Foxg1 and Pax6 expression domains were reduced in the Hesx1^Cre/– mutant anterior forebrain compared with Hesx1^Cre/+ or Hesx1^+/+ controls. In the microarray (full data can be found at ArrayExpress with ID: E-MEXP-2586), levels of Foxg1 and Pax6 expression were found to be 1.86-fold and 1.81-fold higher, respectively, in the Hesx1^eGFP/+ heterozygous relative to the Hesx1^eGFP/eGFP homozygous anterior forebrain precursors. Conversely, the expression domains of Wnt1, Wnt3a, Foxd3 and Pax3 were increased in the Hesx1 null mutant anterior forebrain when visualised by wholemount in situ hybridisation (Andoniadou et al., 2007), and this is recapitulated in the microarray as they showed a 1.6-, 3.2-, 1.61- and 1.69-fold increase in the Hesx1^eGFP/eGFP anterior forebrain cells, respectively. Finally, Hesx1 itself, as expected, showed substantially higher (9.2-fold) expression in the Hesx1^eGFP/+ heterozygous anterior forebrain precursors, ranking as the highest statistically significant differentially expressed gene in the heterozygous sample, corroborating the purity of the mutant cell population and the robustness of the microarray data.

Following an unbiased approach, stringent statistical measures and multiple testing correction, we revealed 55 genes that were expressed at significantly higher levels in the Hesx1^eGFP/eGFP homozygous null mutant anterior forebrain progenitors and 18 genes that were expressed at significantly lower levels (Table 5). This is line with previous research suggesting that HESX1 normally functions as a transcriptional repressor (Carvalho et al., 2010; Dasen et al., 2001; Ermakova et al., 1999; Ermakova et al., 2007). Validation by qRT-PCR on independent samples was carried out for a selection of genes from this shortlist and all 19 genes queried displayed the same trend observed in the microarray (supplementary material Fig. S6). Among the upregulated genes, a significant proportion (20%) were associated with the Wnt pathway, including Tnfrsf19, Dixdc1, Sp5, Apcdd1 and Tnik. A further 46% were associated with processes positively regulated by enhancement of Wnt signalling, namely neural crest specification (16%; e.g. Sox10, Twist1, Ednra, Ednrb) and neural differentiation (30%; Neurog1, Ngfr, Mef2c, Nr2f1, Nrp2).

Certain genes that are known to be upregulated in the forebrain of the Hesx1-deficient embryos, as judged by in situ hybridisation, e.g. the Wnt target Axin2 (Andoniadou et al., 2007), did not pass the strict criteria that we set for statistically significant differences so we also analysed all genes by their Gene Ontology annotations and grouped them by involvement in specific pathways, with differences of 1.6-fold as cut-off. Key components, main effectors and downstream target genes of the major morphogenetic pathways dependent on SHH, FGF or BMP/TGFβ were largely unaffected (supplementary material Tables S3-S5). By contrast, multiple components of the canonical Wnt signalling pathway were upregulated in Hesx1^eGFP/eGFP cells. When subdivided into receptors, ligands, intracellular components, negative regulators and target genes, it became clear that a large number of Wnt targets were expressed at higher levels in the absence of the repressor HESX1 (15 out of 37 genes queried, over 1.6-fold upregulated; supplementary material Table S6). The levels of Wnt receptors and Wnt ligands remained unchanged, except for an increase in the expression of Wnt1, Wnt3a and Wnt6 (Table 5; supplementary material Table S7). Notably, expression of Ctnnb1 was unaffected in Hesx1^eGFP/eGFP cells, as was the expression of genes encoding components of the destruction complex, such as Axin1, Gsk3b, Csk1a1 and Apc, demonstrating that the pathway was not affected at the level of regulation of Ctnnb1 expression or of components affecting degradation of β-catenin protein (supplementary material Table S8).

Together, this gene expression analysis demonstrates that the absence of HESX1 leads to the ectopic activation of numerous Wnt target genes in anterior forebrain progenitors.

In this study, we provide novel genetic and molecular data demonstrating that the homeobox gene Hesx1 antagonises the activation of Wnt/β-catenin signalling in early forebrain progenitors in zebrafish and mouse embryos. In addition, using genetic approaches we reveal a previously unidentified requirement for Tcf3 in these progenitors, where it interacts with Hesx1 to promote anterior character.

Hesx1 has previously been shown to play an essential role during normal forebrain and pituitary development in Xenopus, mice and humans. Here, we show that hesx1 zebrafish morphants display forebrain defects in the sensitised hdl background that are reminiscent of those observed in Hesx1-deficient mouse embryos. These defects are rescued by injection of murine Hesx1 mRNA, suggesting a functional conservation in both species. Significant amino acid sequence homology between Hesx1 proteins of zebrafish and mouse is restricted to the homeobox (DNA-binding domain) at the C-terminus and the engrailed homology 1 (eh1) domain at the N-terminus, which interacts with Groucho/TLE members to mediate transcriptional repression (Carvalho et al., 2010; Dasen et al., 2001). This suggests that the main molecular function of Hesx1/Anf family members is to act as transcriptional repressors.

Members of the TCF/LEF family, including TCF3, activate transcription of Wnt target genes upon association with β-catenin. They are also able to repress genes through association with Groucho/TLE1 co-repressors, an interaction that displaces their association with β-catenin. In the zebrafish embryo, loss of Tcf3a repressor activity was shown to be solely responsible for the anterior truncations displayed in the hdl mutants. Injection of mRNA encoding a truncated form of Tcf3a that lacks the N-terminal β-catenin-interacting domain was capable of rescuing the mutant phenotype through its repressor function. Furthermore, overexpression of the DNA-binding domain fused to the engrailed repressor domain was also able to rescue the hdl anterior defects; however, when fused to the VP16 activator domain it not only failed to rescue this phenotype but also induced forebrain truncations in wild-type embryos (Kim et al., 2000).

Together, these experiments and our MO injections on a sensitised hdl background suggest that there is a functional interaction between Hesx1 and Tcf3a factors that prevents the expression of Wnt targets in the zebrafish embryo. Therefore, the lack of a phenotype in the hesx1 zebrafish morphant could be a consequence of genetic redundancy, whereby Tcf3 factors compensate for the lack of Hesx1.

In the zebrafish embryo, Tcf3 acts independently of β-catenin for normal forebrain development by maintaining the repression of Wnt/β-catenin target genes. Our data extend this analysis and provide evidence that Tcf3 plays this role specifically within the forebrain neuroectoderm. Compound embryos carrying a distinct gene dosage of Hesx1 and Tcf3 show telencephalic and eye defects that are similar to those observed in single-mutant embryos deficient for either Hesx1 or Tcf3. This genetic interaction suggests a requirement not only for TCF3 but also for HESX1 in the inhibition of Wnt/β-catenin target expression in anterior forebrain precursors. We cannot rule out a weak interaction in the axial mesendoderm, but our Dkk1 and Shh analyses suggest that this tissue is unlikely to be affected.

Confirming a novel function of HESX1 in antagonising Wnt signalling, we show that the Hesx1^–/– forebrain defects are partially rescued and forebrain patterning improved in Hesx1^Cre/–;Ctnnb1^LOF/– embryos, in which aberrant Wnt signalling is prevented specifically in anterior forebrain precursors. This illustrates that the forebrain abnormalities in Hesx1-deficient mutants are indeed caused by an ectopic response to this pathway. Deletion of Ctnnb1 in Foxg1-Cre;Ctnnb1^LOF/– embryos has previously been shown to cause forebrain defects (Junghans et al., 2005; Wang et al., 2010), which result from an increase in apoptosis following disruption of structural integrity due to the absence of β-catenin in the adherence junctions of neuroepithelial cells (Junghans et al., 2005) as well as the loss of Fgf8 expression at the ANR (Paek et al., 2011; Wang et al., 2010). By contrast, we did not observe any forebrain phenotype in Hesx1^Cre/+;Ctnnb1^LOF/– embryos. This is possibly due to the different expression patterns of Foxg1 and Hesx1, or is a potential additive effect due to Foxg1 haploinsufficiency in the Foxg1-Cre knock-in line used, as even heterozygous animals have telencephalic defects (Eagleson et al., 2007).

Analysis of double-heterozygous embryos for Hesx1 and another Wnt antagonist, Six3, have also revealed forebrain defects comparable to those of Hesx1;Tcf3 compound mutant embryos (supplementary material Fig. S7) (Gaston-Massuet et al., 2008). Similar to the demonstration that Hesx1 can antagonise Wnt signalling in the zebrafish forebrain, mouse Six3 mRNA is able to rescue the hdl phenotype through antagonising Wnt signalling (Lagutin et al., 2003).

Finally, gene profiling analysis revealed a significant enhancement in the expression of genes relevant to Wnt signalling in the Hesx1-deficient anterior forebrain precursors relative to Hesx1^eGFP/+ heterozygous controls. The increase in expression of several target genes of the Wnt/β-catenin pathway in the Hesx1^eGFP/eGFP anterior forebrain cells demonstrates the ectopic activation of this pathway in the absence of Hesx1 in a tissue that would normally be unresponsive to Wnt signals. Indeed, in the Hesx1^eGFP/eGFP population, we observe an increase in the Wnt effectors Lef1 (1.7-fold) and Tcf1 (Tcf7, 1.6-fold), which are not normally expressed in the anterior forebrain. The ectopic expression of Sp5 in the forebrain of Hesx1^Cre/+;Tcf3^fl/– embryos leads to the notion of a similar underlying defect in these mutants. This supports the notion that Hesx1 may act, in concert with Tcf3 and Six3, as a negative regulator of the Wnt pathway in the anterior forebrain, furthering our understanding of the mechanisms required to establish forebrain identity. In addition, the microarray data provide a valuable resource because they define the normal molecular signature of early anterior forebrain precursors. This information can be used for comparative studies with other mouse mutants or to assess the efficiency of protocols for in vitro differentiation of stem cell lines into neurons.

The variability in the forebrain defects observed in compound mutants also exists in human patients with mutations in HESX1. To date, more than 15 mutations have been identified in human HESX1 in association with variable degrees of forebrain and pituitary defects (Kelberman et al., 2009). Our mouse research opens the possibility for mutations in genes with synergistic action, such as TCF3, to be candidates for modifying these phenotypes.

We thank Professor Andrew Copp for critical reading of the manuscript. This work was carried out with the support of the UCL Institute of Child Health and Great Ormond Street Hospital Flow Cytometry Core Facility, UCL Genomics, the ICH Embryonic Stem Cell/Chimera Production Facility and UCL Biological Services Unit.