Head formation: OTX2 regulates Dkk1 and Lhx1 activity in the anterior mesendoderm

The gene orthodenticle homologue 2 (Otx2) encodes a paired-type homeobox transcription factor that controls the expression of target genes by binding to regulatory regions containing the YTAATNN motifs (Kimura-Yoshida et al., 2005; Chatelain et al., 2006). In the mouse, Otx2 is expressed in the inner cell mass of the blastocyst (at embryonic day E3.5) and in the visceral endoderm of the E5.5 pre-gastrulation embryo (Kimura et al., 2001). It is also initially widely expressed in the epiblast, but becomes progressively regionalised during gastrulation to an anterior domain of the ectoderm (anterior neuroectoderm, ANE) that contains the precursor cells of the prosencephalon and the mesencephalon of the embryonic brain (Tam, 1989; Ang et al., 1994; Cajal et al., 2012).

Loss of Otx2 in the mouse impedes morphogenetic tissue movement in the pre-gastrulation embryo, arrests germ layer formation during gastrulation and disrupts the morphogenesis of the embryonic head (Acampora et al., 1995; Matsuo et al., 1995; Ang et al., 1996). In Otx2^−/− mutants, the precursor cells for the anterior visceral endoderm (AVE) fail to translocate from the distal site to the anterior region of the pre-gastrulation stage embryo. The absence of the AVE causes the loss of the source of morphogenetic signals for inducing neural differentiation of the ANE (Kimura et al., 2000). Lineage-tracing experiments have shown that Otx2 activity in the visceral endoderm is required for navigating the cell movement (Kimura et al., 2000; Perea-Gomez et al., 2001).

Among the signalling pathway factors that are expressed in the visceral endoderm (Pfister et al., 2007) is Dickkopf homologue 1 (DKK1), an antagonist of WNT signalling (Kemp et al., 2005; Lewis et al., 2007; Fossat et al., 2012) and a downstream target of OTX2 (Kimura-Yoshida et al., 2005). Dkk1 expression in the visceral endoderm is abolished in the Otx2^−/− embryos (Zakin et al., 2000; Perea-Gomez et al., 2001). In transgenic Otx2^Dkk1/Dkk1 embryos, expression of Dkk1 from the Otx2 locus can rescue the migration defects of Otx2-deficient visceral endoderm cells. Head development, however, is still defective, indicating that positioning of the visceral endoderm is insufficient to reconstitute the genetic or signalling activity required for head formation (Kimura-Yoshida et al., 2005). Otx2 function can be substituted by another Otx gene, Otx1, or its fly homologue otd. Chimeric mice or Otx2^Otx1/Otx1 and Otx2^otd/otd transgenic mice that respectively express Otx2, Otx1 or otd only in the visceral endoderm display normal cell migration in the visceral endoderm and the induction of the ANE, but still fail to form a proper brain (Acampora et al., 1998, 2001; Rhinn et al., 1998). These findings suggest that Otx2 function is required in other embryonic tissue besides the visceral endoderm for maintaining the differentiation potential of the ANE.

Chimera experiments have shown that OTX2 acts cell-autonomously to maintain the expression of the anterior neuronal markers Hesx1, Wnt1, Cdh4 and Efna2 in the ANE (Rhinn et al., 1999), whereas inactivating Otx2 in epiblast stem cells impairs the differentiation of anterior neural tissues (Iwafuchi-Doi et al., 2012; Acampora et al., 2013). These results highlight the intrinsic requirement of Otx2 activity in the ANE for differentiation. The layer of endoderm cells and the midline structure (the anterior mesendoderm, AME) that are associated with the ANE are both derived from the epiblast during gastrulation. The AME expresses signalling pathway genes and its integrity is essential for the development of the anterior neural structures (Camus et al., 2000; Hallonet et al., 2002; Robb and Tam, 2004; Pfister et al., 2007). Otx2 is also expressed in the AME but its tissue-specific function in the induction or maintenance of the ANE has not been tested directly.

In the present study, we have examined the requirement of Otx2 in the AME by Cre-mediated ablation of the gene activity in Otx2 AME-conditional knockout (Otx2-ameCKO) mutant embryos. Our results show that Otx2 is required in the AME for the induction of anterior neural tissues and head formation. Loss of Otx2 in the AME is accompanied by the downregulation of Dkk1 and Lhx1. Results of RT-qPCR, ChIP-qPCR and luciferase assays indicate that these two genes are likely to be direct transcriptional targets of OTX2 in the AME. The analysis of compound mutant embryos shows that they both act synergistically with Otx2 in this structure for head formation.

To analyse the specific requirement of Otx2 in the AME, we used the Foxa2-mcm mouse line that expresses tamoxifen-inducible MerCreMer recombinase [Foxa2^{tm2(Cre/Esr1*)Moon} – Mouse Genome Informatics] in Foxa2-expressing AME (Park et al., 2008). The expression pattern of R26R Cre-reporter (Soriano, 1999) in Foxa2^+/mcm;Rosa26^+/R26R embryos treated with tamoxifen at E6.5 and harvested 30 h later revealed the presence of X-Gal-stained cells in the endoderm layer, predominantly in the anterior definitive endoderm and the axial AME (Fig. 1A). Otx2^flox/flox;Foxa2^mcm/mcm mice were crossed with Otx2^+/− mice to produce Otx2^+/flox;Foxa2^+/mcm and Otx2^flox/−;Foxa2^+/mcm embryos. These embryos were treated with tamoxifen or mock vehicle once at E6.5 and collected 30 h after treatment. Histology of the embryo showed that Otx2 expression was reduced in the midline tissue of the tamoxifen-treated Otx2^flox/−;Foxa2^+/mcm embryos (the Otx2-ameCKO embryos) compared with vehicle-treated (mock control) Otx2^flox/−;Foxa2^+/mcm embryos (Fig. 1B). In addition, the epithelial cells in the midline of the Otx2-ameCKO embryo were shorter than the columnar AME cells of the control embryo (Fig. 1B).

Whereas Foxa2-mcm activity was widely expressed in the endoderm and the AME in the anterior region of the embryo after tamoxifen treatment at E6.5 (Fig. 1A), the impact of the ablation of Otx2 is likely to be limited to the AME, based on the following considerations: (a) Otx2 expression is confined to the tissues close to the midline, i.e. the AME and the immediately adjacent endoderm cells, thus suggesting that the ablation of Otx2 in the non-expressing cells would not have any impact on the phenotype; (b) the descendants of the AVE that are present in the E6.5 embryo at the time of tamoxifen treatment are mostly found in the extraembryonic yolk sac (Thomas and Beddington, 1996; Shimono and Behringer, 2003). That practically no lacZ-positive cells were present in the yolk sac endoderm of the Foxa2^+/mcm;Rosa26^+/R26R embryos treated with tamoxifen at E6.5 (Fig. 1A) suggests that the Cre activity was not activated in the AVE, which should still harbour an undeleted Otx2^flox allele. However, our results do not exclude the possibility that some AVE descendants in the anterior endoderm (Kwon et al., 2008) contain a Cre-mediated ablation of the Otx2 gene. Nevertheless, it is likely that the presence of a small subset of Otx2-deficient AVE descendants (given the mosaic pattern of Cre activity) has no significant functional consequences that contribute to the mutant phenotype.

At E9.5, the Otx2-ameCKO embryos displayed a fully penetrant head truncation phenotype (in a range between category II and V; see Materials and Methods) (Table 1; Fig. 1C). Fourteen percent of mock control Otx2^flox/−;Foxa2^+/mcm embryos showed a mild truncation phenotype (category II) (Table 1) that is reminiscent of the hypomorphic heterozygous Otx2^+/− embryo (Acampora et al., 1995; Fossat et al., 2006). A similar frequency of mild truncation (category II) was found in the tamoxifen-treated Otx2^+/flox;Foxa2^+/mcm embryos but not in the mock control with the same genotype (Table 1). Analysis of molecular markers revealed a reduced dorsal and ventral rostral telencephalon (Fgf8, Six3) and dorsal diencephalon (Tcf4) in the Otx2-ameCKO embryos (Fig. 1D). Otx2 activity in the AME is therefore essential for head formation.

In the Otx2-ameCKO embryos, Shh-expressing cells were present in the anterior midline that resembled the AME of the mock control embryo (Fig. 2A), indicating that the loss of Otx2 in the Foxa2-expressing tissue does not affect AME formation. Given that Dkk1 is a known target of OTX2 (Kimura-Yoshida et al., 2005) and that it is required in the AME for head development (Mukhopadhyay et al., 2001; Lewis et al., 2007), we examined Dkk1 expression in the Otx2-ameCKO embryos and observed that its expression was reduced in the AME of these mutants (Fig. 2B).

Loss of Lhx1, which encodes the LIM-homeobox 1 transcription factor, also leads to severe head truncation (Shawlot and Behringer, 1995). Lhx1 is required in both the visceral endoderm and the epiblast for head formation. Chimeric embryos that have lost Lhx1 activity in the visceral endoderm or the epiblast phenocopy the head truncation of the Otx2-ameCKO embryos (Shawlot et al., 1999). Consistent with a contribution by Lhx1 to the genesis of the mutant phenotype, Lhx1 expression was downregulated in the midline cells in the endoderm layer, but not in the definitive endoderm or the mesoderm of the Otx2-ameCKO embryo (Fig. 2B). These results suggest that the reduction or loss of Lhx1 and Dkk1 activity in the AME underpins the impact of loss of Otx2 function on forebrain and head formation.

To investigate whether OTX2 can regulate Dkk1 and Lhx1, their expression was assessed in mouse P19 embryonal carcinoma cells transfected with plasmid expressing wild-type OTX2 or a mutant form of OTX2 (hereafter referred to as OTX2-3pm) that is unable to bind DNA due to three mismatch point mutations in the homeodomain coding sequence (Chatelain et al., 2006). Both the wild type and the variant form of OTX2 were tagged with GFP and were expressed in the nucleus (Fig. 3A), as previously reported (Chatelain et al., 2006). In the transfected P19 cells, Dkk1 and Lhx1 expression were upregulated by OTX2 but not by OTX2-3pm (Fig. 3B). These results show that OTX2 can activate Lhx1 and Dkk1 expression and that this is dependent on the ability of OTX2 to bind DNA.

OTX2 protein binds YTAATNN motifs (Kimura-Yoshida et al., 2005; Chatelain et al., 2006) and is known to directly control Dkk1 expression by binding the H1 regulatory region located ∼500 bp upstream of the Dkk1 START codon (Kimura-Yoshida et al., 2005). No OTX2 regulatory regions have been reported for the Lhx1 locus. We therefore performed a multi-species comparative analysis to search for conserved regions containing at least one OTX2-binding motif and located within 50 kb upstream and 15 kb downstream of the START codon of Lhx1. Three such regions were found: two located 30 kb (+30R) and 1.5 kb (+1.5R) upstream, and one located 6.5 kb (−6.5R) downstream of the START codon (supplementary material Figs S1-S3). The binding of OTX2 to these regions was tested by chromatin immunoprecipitation (ChIP)-qPCR experiments using an anti-GFP antibody and P19 cells expressing Otx2-gfp or Otx2-3pm-gfp. +1.5R and −6.5R regions were significantly enriched in chromatin precipitated with OTX2 when compared with chromatin precipitated with OTX2-3pm (Fig. 3D). The same was observed for the +30R region, although the binding appeared to be weaker. We could also confirm the binding of OTX2 to the H1 region of Dkk1. No enrichment was observed for a region of the genome that has no OTX2-binding motif (NOBR).

We tested whether these conserved regions could act as enhancers to mediate the activation of gene expression by OTX2. These sequences were cloned into a reporter plasmid that carries a minimal promoter and a gene encoding the firefly luciferase protein. Elevated firefly luciferase activity (compared with the mock control) was detected when co-transfected with an Otx2 expression vector and the reporter plasmid containing the Dkk1 H1 region, and with the Lhx1 +1.5R and −6.5R regions, but not with the Lhx1 +30R region or the DNA fragment without OTX2-binding motif (NOBR) (Fig. 3E). Luciferase activity was not different from the mock control when P19 cells were co-transfected with an Otx2-3pm expression vector (Fig. 3E). These results show that both the Dkk1 H1 region and the Lhx1 +1.5R and −6.5R regions are bound by OTX2 to activate gene expression. Interestingly, ChIP-sequencing data of the ENCODE project support the notion that regions equivalent to +1.5R and −6.5R regions in human LHX1 are regulatory (Wang et al., 2012).

Loss of Otx2 produces a phenocopy of the Lhx1-null and Dkk1-null head truncation phenotype (Acampora et al., 1995; Matsuo et al., 1995; Shawlot and Behringer, 1995; Ang et al., 1996; Mukhopadhyay et al., 2001). Otx2, Lhx1 and Dkk1 display overlapping domain of expression in the AME during gastrulation (Fig. 4A). The downregulation of Lhx1 and Dkk1 in the AME of Otx2-ameCKO embryos raised the possibility that Otx2 acts synergistically with Lhx1 and Dkk1 in facilitating head development. We tested this hypothesis by analysing the phenotype of AME-specific compound Otx2;Lhx1 and Otx2;Dkk1 heterozygous mutant embryos. This was accomplished by generating Otx2^+/flox;Lhx1^+/−;Foxa2^+/mcm (compound Otx2;Lhx1) and Otx2^+/flox;Dkk1^+/−;Foxa2^+/mcm (compound Otx2;Dkk1) mutant embryos from a cross between Otx2^flox/flox;Foxa2^mcm/mcm and Lhx1^+/− mice or Dkk1^+/− mice, respectively. Induction of MerCreMer recombinase was achieved by tamoxifen administration at E6.5, as for the Otx2-ameCKO embryos. The AME of these mutant embryos were populated by Otx2^+/−;Lhx1^+/− cells (in compound Otx2;Lhx1 mutants) or by Otx2^+/−;Dkk1^+/− cells (in compound Otx2;Dkk1 mutants). In the compound Otx2;Lhx1 embryos, the mutants displayed more severe head truncation at a higher frequency than mutant embryos with mock injection (Fig. 4B; Table 2), as well as Otx2^+/flox;Lhx1^+/+;Foxa2^+/mcm embryos generated from the same cross (tamoxifen- and mock-injected; Fig. 4B; Table 2). Similarly, the compound Otx2;Dkk1 embryos also displayed a more severe truncated head phenotype at a higher frequency than mutant embryos with mock injection or tamoxifen- and mock-treated Otx2^+/flox;Dkk1^+/+;Foxa2^+/mcm embryos (Fig. 4C; Table 3). Taken together, these results show that Otx2 acts synergistically with Lhx1 and Dkk1 specifically in the AME for head formation.

Results of this study have highlighted a crucial role for Otx2 in the AME in promoting head and forebrain formation, and that OTX2 function might be mediated by regulating the activity of two downstream genes, Lhx1 and Dkk1, that are also essential for head morphogenesis. These two genes are likely to be the transcriptional targets of OTX2 that activates their expression by binding conserved enhancer regions.

Otx2 is expressed at multiple sites in the early embryo, such as the visceral endoderm, the epiblast and the epiblast-derived ANE and AME, and might have tissue-specific functions at these sites during the formation of the head primordium (Ang et al., 1994; Kimura et al., 2000). In the visceral endoderm of the pre-gastrulation stage embryo, Otx2 is required for navigating the movement of cells towards the prospective anterior region (Acampora et al., 1998, 2001; Rhinn et al., 1998; Kimura et al., 2000; Perea-Gomez et al., 2001). The directional translocation of cells is instrumental for the establishment of the AVE. Loss of Otx2 function impedes the movement of the visceral endoderm, which can be overcome by expressing Dkk1 from the mutant Otx2 locus (Kimura-Yoshida et al., 2005). Restoring the AVE through enhancing Dkk1 activity in the visceral endoderm, however, does not rescue the Otx2-deficient embryo from developing head truncation defects. In a similar context, chimeric embryos comprising Otx2-expressing visceral endoderm and Otx2-deficient epiblast (Rhinn et al., 1998), and conditional mutant embryos with Otx2 knocked out in Otx2-expressing epiblast-derivatives by inducible Otx2-CreER^T2 activity (Fossat et al., 2006) also develop head truncation defects. By regulating expression of homeobox and cell adhesion genes Otx2 has both cell-autonomous and non-autonomous functions in the ANE. Chimera embryos containing both Otx2 mutant and wild-type cells in the ANE show reduced Hesx1, Wnt1, Cdh4 and Efna2 expression in mutant cells, which eventually enter into apoptosis (Rhinn et al., 1999). Otx2 is also required for the dorso-ventral patterning of the brain (Puelles et al., 2003) and to delineate the mid-hindbrain boundary by antagonising Gbx2-expressing cells (Li and Joyner, 2001; Martinez-Barbera et al., 2001). These findings therefore indicate a separate requirement of Otx2 activity in the epiblast and its derivatives, probably for the induction and maintenance of the neural potency of the anterior ectoderm and for brain morphogenesis. However, it is not known in which epiblast tissue compartment Otx2 plays an essential role in the early phase of head development.

By targeting the ablation of Otx2 to the Foxa2-expressing cells in the gastrula-stage embryo, Otx2 expression has been specifically eliminated in the AME. The conditional mutant Otx2-ameCKO embryos display head truncation defect, principally a reduced forebrain and loss of the facial primordia, although not as severe as the complete truncation at the mid-hindbrain junction of the Otx2-null embryo (Acampora et al., 1995; Matsuo et al., 1995; Ang et al., 1996). The manifestation of a mutant phenotype is not due to the failure to form the anterior axial mesendoderm, but is associated with impaired morphogenetic activity, which is revealed by the downregulation of genes with key roles in head formation. Otx2 is therefore required in a multi-phasic manner, first in the extraembryonic tissue for establishing the signalling source for tissue patterning, and later in the epiblast-derived tissues, where it plays a role in the AME for the maintenance of the anterior characteristics of the overlying ectoderm (this study), and its activity in the ANE drives the differentiation of the anterior neural progenitors.

The Lhx1 gene that is downregulated in the AME of the Otx2-ameCKO embryo is similarly important for head formation and also displays several modes of functional requirement in head development. Loss of Lhx1 leads to severe head truncation as in the Otx2-null mutant (Shawlot and Behringer, 1995). Lhx1-null embryos display abnormal shape due to disruption in germ layer morphogenesis and fail to form a proper head (Shawlot and Behringer, 1995; Shawlot et al., 1999). In the visceral endoderm, Lhx1 is required to maintain angiomotin activity for the movement of the AVE at the onset of gastrulation (Shimono and Behringer, 1999,, 2003). Lhx1 is also required in epiblast-derived tissues for head formation (Shawlot et al., 1999). Both the chimeric embryo with Lhx1-active visceral endoderm and Lhx1-deficient epiblast, and the conditional mutant embryo with no Lhx1 activity in the epiblast-derived tissues develop head truncation defects (Shawlot et al., 1999; Kwan and Behringer, 2002; Tanaka et al., 2010). In the present study, the synergistic effect of the combined loss of function of Otx2 and Lhx1 on the manifestation of the head phenotype and the effector-target relationship of Otx2 and Lhx1 are consistent with their co-requirement for head formation. Our results suggest that both gene products might be required in the AME for head formation, namely OTX2 by activating Lhx1 expression and LHX1 by controlling target genes crucial for head formation that still need to be identified.

Some clues of the Otx2 downstream activity that might have a role in head formation can be obtained from the study of its other target, Dkk1. DKK1 is a secreted molecule that inhibits WNT signalling for head formation (Fossat et al., 2012). Genetic experiments in mice demonstrated that an elevated level of WNT signalling activity impairs anterior development and that the increase in WNT signalling and the impairment are exacerbated by the absence of Dkk1 (Fossat et al., 2011,, 2012; Arkell and Tam, 2012; Arkell et al., 2013). The AVE is a source of secreted antagonists (Dkk1, Sfrp1, Sfrp5 and Cer1) for modulating WNT signalling. However, lacking Dkk1 activity only in the visceral endoderm has no impact on head formation (Mukhopadhyay et al., 2001). In the gastrulating embryo, Dkk1 expression is restricted to the AME and the head truncation phenotype of Dkk1^−/− embryos is likely to be due to the absence of Dkk1 in this structure (Mukhopadhyay et al., 2001). Dkk1 is a target of OTX2 in the AVE (Kimura-Yoshida et al., 2005). Here, we demonstrate that conditional ablation of Otx2 in the AME leads to the downregulation of Dkk1, which also supports a direct regulation of Dkk1 expression by OTX2 in the AME. Furthermore, we show a synergistic interaction between Otx2 and Dkk1 in the AME. OTX2 might therefore function through the regulation of Dkk1 expression in the AME to modulate the level of WNT signalling perceived by the ANE during the development of the anterior neural structures.

Our study has therefore revealed a functional hierarchy of Otx2-Dkk1 activity in the AME and that one function of Otx2 is likely to be modulating the level of canonical WNT signalling activity for neural induction and the morphogenesis of the head. Whereas our findings have also established an Otx2-Lhx1 transcriptional regulatory relationship, the downstream activity of the Otx2-Lhx1 interaction in controlling head formation is presently unknown and awaits further investigation.

Otx2^flox/flox and Otx2^+/− mice (Fossat et al., 2006), Foxa2^mcm/mcm mice (Park et al., 2008), Lhx1^+/− mice (Kwan and Behringer, 2002), Dkk1^+/− mice (Mukhopadhyay et al., 2001) and Rosa26^+/R26R mice (Soriano, 1999) were used for the genetic experiments in this study. Genotyping by PCR followed established protocols and was performed on DNA extracted from tail tissues of newborn mice or the yolk sac of embryos. The breeding strategies for the production of compound mutants as well as the number of embryos of each genotype that were analysed for head defects are described in Tables 1–3. All animal experimentation was approved by the Animal Ethics Committee of the Children's Medical Research Institute and of the Children's Hospital at Westmead, Sydney, Australia.

Embryos were collected from gravid mice at the required gestational age or at specific time points after tamoxifen or mock treatment. Tamoxifen (Sigma-Aldrich) diluted at 10 mg/ml in canola oil was administered to gravid females by intraperitoneal injection at 100 µl (1 mg) per 20 g of body weight (the day of vaginal plug was taken as E0.5). For mock control, the same volume of canola oil without tamoxifen was administered. Embryos were staged according to criteria established by Downs and Davies (1993) or by somite number.

The morphology of mutant embryos was compared with stage-matched control embryos with specific attention to the size, the number of somites and the morphology of the head. The degree of head truncation was determined by measuring the head size, which was expressed as a ratio of the length of the dorsal margin of the silhouette of the head from the junction of the maxillary prominence and the base of the first branchial arch to the anterior border of the otic vesicles to the diameter of the otic vesicle (for a description of the normalisation of the measurement to account for the developmental status of the embryo see Lewis et al., 2007). Embryos were assigned to one of five categories based on the size of the fore- and midbrain: [I] Normal size (0% reduction), [II] ≤25% reduction, [III] 26-75% reduction, [IV] >75% reduction, [V] complete absence. As an example, the head size of class I embryos (normal size) was 15.37±0.42 units (n=6), compared with that of class II (the mildest reduction at ≤25%) size of 12.81±1.09 units (n=4), which is significantly different at P<0.0007 by two-tailed t-test.

In situ hybridisation and X-Gal staining were performed as previously described (Tam and Steiner, 1999; Fossat et al., 2007, 2011; Lewis et al., 2007). For histology analysis, embryos were embedded in paraffin, sectioned and counterstained with nuclear Fast Red and Eosin according to standard protocols (Tam and Steiner, 1999; Lewis et al., 2007). At least three specimens of each genotype were analysed for each marker tested.

Expression vectors: pCMV-GFP, pCMV-Otx2-GFP and pCMV-Otx2-3pm-GFP plasmids have been described (Chatelain et al., 2006; Fossat et al., 2014).

Vectors for luciferase assay: regulatory elements were amplified by PCR from mouse genomic DNA with primers described below and cloned into the XhoI site or between the NheI and the KpnI sites of the pGL3-promoter plasmid (Promega). The H1 region of Dkk1 (Kimura-Yoshida et al., 2005) was amplified with 5′-ATCGCTCGAGCGAGGTTTGATTTGGGATCA-3′ and 5′-ATCGCTCGAGTATGAGTACCCTCGCTTTCT-3′ primers. The region 30 kb upstream of Lhx1 ATG (+30R; 29,827-30,053 bp upstream of Lhx1 START codon) was amplified with 5′-ATCGGCTAGCCCTG-TGCTCTCTTGTCCAT-3′ and 5′-ATCGGCTAGCAGGCCTGACATTT-CCAACAT-3′ primers. The region 1.5 kb upstream of Lhx1 ATG (+1.5R; 1581-1877 bp upstream of Lhx1 START codon) was amplified with 5′-ATCGGGTACCAAACAGACAGAAGGCGAGAT-3′ and 5′-ATCGG-CTAGCTCCGTTTCGGCATCAAGGA-3′ primers. The region 6.5 kb downstream of Lhx1 ATG (−6.5R; 6591-6809 bp downstream of Lhx1 START codon) was amplified with 5′-ATCGGGTACCGCTTGTGG-ATGTCAGATACT-3′ and 5′-ATCGGCTAGCTCTGGGGAACTTCAA-ACGTA-3′ primers. A region of the genome not bound by Otx2 (NOBR; −1402 bp to −1660 bp downstream of Hesx1 START codon) was amplified using 5′-ATCGGGTACCTTGGCTTCTCTTCCAGTTGT-3′ and 5′-ATCGGCTAGCCTGTATACACCTGTGAAGGT-3′ primers.

Mouse embryonic carcinoma P19 cells and 3T3 fibroblasts were transfected using Fugene 6 (Roche) with pCMV-GFP, pCMV-Otx2-GFP or pCMV-Otx2-3pm-GFP. At 24 h after transfection, cells were washed and fixed for immunostaining or collected and snap-frozen for ChIP-qPCR experiments. For RT-qPCR analysis, at least 250,000 GFP-expressing cells were isolated by fluorescence-activated cell sorting (FACS) before being snap-frozen. For the luciferase assay, 3T3 cells were co-transfected with the pGL3-promoter constructs and with pRL Renilla luciferase control reporter vectors (Promega).

Cells were washed with PBS, incubated for 5 min in 4% paraformaldehyde (PFA) with 0.5% Triton X-100, then for 20 min in 4% PFA at room temperature. After fixation, cells were washed with PBS, incubated for 15 min at room temperature with CAS-Block (Invitrogen), then for 40 min at 4°C with anti-GFP antibody (1:300; A11122, Invitrogen) diluted in CAS-Block. The following day, cells were washed with PBS and incubated for 40 min at room temperature with Alexa Fluor 488 anti-rabbit antibody (1:300) and 4′,6-diamidino-2-phenylindole (DAPI; Sigma-Aldrich; 1:1000) diluted in PBS. Cells were finally washed with PBS and embedded in Fluoromount-G (Invitrogen) for microscopy analysis.

RNA from three independent transfections was extracted using an RNeasy Micro Kit (Qiagen). cDNA was synthesised from 100 ng RNA using the Superscript III First Strand System (Invitrogen). Quantitative PCR was performed in triplicate from 1:5 dilution of cDNA of each sample using the Rotorgene 6000 thermal cycler (Corbett Research), SYBR Green I (Molecular Probes) and Platinum Taq DNA Polymerase (Invitrogen). Primers used for assaying β-actin and Dkk1 expression were previously described (Fossat et al., 2011). The primers 5′-AAGCAACTGGA-GACGTTGAA-3′ and 5′-CTAGCGCGCTTAGCTGTTT-3′ were used to assay Lhx1 expression. PCR conditions (Fossat et al., 2011) were similar for all three genes.

Cells were collected from three independent transfections. The luciferase assay was performed using the Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer's instructions. Firefly and Renilla luciferase activities were detected by a luminometer [Turner Designs TD-20/20 luminometer; dual-luciferase Renilla (DLR) mode; time-measurement delay 2 s; integrate time 6 s]. Relative firefly luciferase activity was measured and normalised relative to Renilla luciferase activity. Furthermore, all measurements were normalised to luciferase activity quantified from an experiment performed with a reporter vector without regulatory regions (pGL3-promoter).

Three million P19 cells were washed with PBS, dissociated with trypsin and centrifuged at 800 g for 5 min. The cell pellet was resuspended and rotated in 1 ml PBS with 1% formaldehyde for 10 min. Fixation was stopped by adding 0.15 M glycine. From this point on, the experiment was performed at 4°C. Cells were washed three times with ice-cold PBS and centrifuged at 800 g for 10 min. The cell pellet was resuspended and incubated for 10 min in 1 ml LB-A buffer [50 mM Hepes-KOH, pH 7.5, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% Igepal, 0.25% Triton X-100, 1× complete protease inhibitor (Roche)]. The lysate was centrifuged at 800 g for 10 min and the cell pellet was resuspended and incubated for 10 min in 1 ml LB-B [10 mM Tris-HCl, pH 8.0, 200 mM NaCl, 1 mM EDTA, pH 8.0, 0.5 mM EGTA, pH 8.0, 1× complete protease inhibitor (Roche)]. The lysate was centrifuged at 800 g for 10 min and the nuclear pellet was resuspended and incubated for 10 min in 1 ml LB-C [50 mM Hepes, pH 7.5, 40 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 0.1% Na-deoxycholate, 0.1% SDS, pH 8.0, 1× complete protease inhibitor (Roche)]. Nuclear lysate was sonicated (seven times: 30 s, 4 cycles) with a Bioruptor (Diagenode). The sonicated sample was centrifuged at 18,000 g for 10 min and 50 µl of the supernatant was collected for input control. The rest of the supernatant (∼950 µl) was rotated with 60 µl protein G Dynabeads (Invitrogen) for 3 h for pre-clearing. The beads were discarded and the pre-cleared lysate was rotated with 5 µg of anti-GFP (A11122, Invitrogen) antibody overnight. In parallel, 40 µl protein G Dynabeads were rotated with blocking buffers (0.5% BSA and 0.5% Tween-20 in PBS without Ca²⁺ and Mg²⁺) overnight for pre-blocking. The next day, pre-cleared antibody-bound chromatin complexes (Ab-chromatin complexes) were rotated with pre-blocked beads for 1.5 h. The Ab-chromatin complex-bound beads were washed each time for 5 min with the following buffers: RIPA buffer [10 mM Tris-HCl, pH 8.0, 1 mM EDTA, pH 8.0, 140 mM NaCl, 1% Triton X-100, 0.1% SDS, 0.1% Na-deoxycholate, 1× complete protease inhibitor (Roche)]; five times with RIPA buffer without protease inhibitor; twice with RIPA-500 buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA, pH 8.0, 500 mM NaCl, 1% Triton X-100, 0.1% Na-deoxycholate); twice with LiCl buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA, pH 8.0, 250 mM LiCl, 0.5% NP-40, 0.5% Na-deoxycholate) and once with TE buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA, pH 8.0). The beads were then resuspended in 50 µl elution buffer (10 mM Tris-HCl, pH 8.0, 5 mM EDTA, pH 8.0, 300 mM NaCl, 0.5% SDS) with 2 µg RNase A (Invitrogen) and incubated for 30 min at 37°C. Proteinase K (2.5 µl, 20 mg/ml; Invitrogen) was added and the reaction incubated overnight at 62°C for reverse cross-linking. The beads were then discarded. A DNA purification step, to extract chromatin fragments, was performed with solid phase reversible immobilisation (SPRI) beads using 2.3 volumes of SPRI beads per volume of sample for 4 min (Blecher-Gonen et al., 2013). Beads were then washed twice with 70% ethanol and air-dried for 4 min. Chromatin was eluted in 30 µl Tris buffer (10 mM). A size-selection step was performed with SPRI beads by first adding 0.6 volumes (18 µl) of SPRI beads per volume of sample and incubation for 4 min at room temperature. Beads binding chromatin fragments larger than 500 bp were then discarded. Subsequently, the chromatin was incubated with 1.8 volumes (36 μl) of SPRI beads for 4 min at room temperature, washed twice with 70% ethanol and air-dried for 4 min. Chromatin was eluted in 30 µl Tris buffer (10 mM).

qPCR was performed from 2 µl immunoprecipitated chromatin or 2 µl input sample, using a Rotorgene 6000 thermal cycler (Corbett Research), SYBR Green I (Molecular Probes) and Platinum Taq DNA Polymerase (Invitrogen). The following primers were used: Dkk1 H1 region, 5′-CGAGGTTTGATTTGGGATCA-3′ and 5′-CACATGAGATCAAAG-TGGCT-3′; Lhx1 +30R region, 5′-CCTGTGCTCTCTTGTCCAT-3′ and 5′-AGGCCTGACATTTCCAACAT-3′; Lhx1 +1.5R region, 5′-AAACA-GACAGAAGGCGAGAT-3′ and 5′-TCCGTTTCGGCATCAAGGA-3′; Lhx1 −6.5R region 5′-GCTTGTGGATGTCAGATACT-3′ and 5′-TCTG-GGGAACTTCAAACGTA-3′; NOBR region +8142-8341 bp upstream of Dkk1 START codon, 5′-AAACTGGCTGCCCATCCAT-3′ and 5′-TCCC-GACAATACTAGGGATA-3′. PCR conditions (Fossat et al., 2011) were similar for all regions. The relative enrichment of a region in the immunoprecipitated sample was determined by normalising the C_t value obtained from the immunoprecipitated sample to the C_t value obtained from the input sample. Relative enrichment was calculated by the following equation: 2^{input Ct–immunoprecipitated Ct}.

Mouse and human genomic sequence were analysed with the University of California, Santa Cruz (UCSC) genomic browser and YASS (Noe and Kucherov, 2005).

We thank Xin Wang for help with flow cytometry, performed in the Flow Cytometry Centre supported by the Westmead Millennium Institute, the NHMRC and the Cancer Institute, NSW, Australia. We thank the CMRI Bioservices Unit for animal care.