Regulation of Otx2 expression and its functions in mouse epiblast and anterior neuroectoderm

The Otx family of paired-type homeobox genes comprises vertebrate homologs of Drosophila head gap gene otd(Simeone et al., 1992). Mammals possess two homologs, Otx1 and Otx2, whereas Xotx4 and otx3 are present in the natural tetraploid genomes of Xenopus and zebrafish; Xotx4 and otx3 are divergent copies of Otx1. Otx2 is expressed in each step and at each site of head development throughout the tetrapod lineage. In mouse, its expression is seen in the epiblast, distal/anterior visceral endoderm,anterior mesendoderm, anterior neuroectoderm, forebrain/midbrain and cephalic mesenchyme. By contrast, Otx1 expression occurs later, in forebrain and midbrain (Simeone et al.,1992; Simeone et al.,1993; Li et al.,1994; Mori et al.,1994; Bally-Cuif et al.,1995; Mercier et al.,1995; Pannese et al.,1995; Kablar et al.,1996; Kimura et al.,1997; Kimura et al.,2000; Kimura et al.,2001). In zebrafish, however, the expression patterns of Otx homologs are divergent. otx1 is expressed in blastoderm prior to the shield stage. In addition, otx1 and otx3, but not otx2, are expressed in the shield, i.e. the fish organizer(Mori et al., 1994). All zebrafish Otx genes are expressed in anterior mesendoderm. otx1 and otx3 expression also precede otx2 expression in anterior neuroectoderm.

Mutant studies in mouse have suggested that Otx2 plays essential roles in head development (Acampora et al.,1995; Acampora et al.,1997; Acampora et al.,1998; Ang et al.,1996; Matsuo et al.,1995; Rhinn et al.,1998; Rhinn et al.,1999; Suda et al.,1996; Suda et al.,1997; Suda et al.,1999; Suda et al.,2001; Tian et al.,2002; Kimura et al.,1997; Kimura et al.,2000; Kimura et al.,2001). However, owing to its expression in multiple sites, it has not to date been possible to distinguish clearly the functions of Otx2 in each site. In addition, information is scarce regarding the upstream factors that regulate Otx2 expression at each developmental site. In light of the essential roles played by Otx genes in head development,it will be important to develop an understanding of how mechanisms governing the regulation of Otx expression in each vertebrate are related and altered.

Previously, we have identified mouse enhancers of Otx2 expression in visceral endoderm (VE), definitive anterior mesendoderm (AME) and cephalic neural crest cells or mesenchyme (CM); the sites located proximal to the Otx2-coding region (Kimura et al., 1997; Kimura et al.,2000). However, no ectodermal enhancers were present in this proximal region. Enhancers of Otx2 expression in epiblast, anterior neuroectoderm and forebrain/midbrain were then sought by BAC transgenesis, and indications were that they existed within the –170 kb 5′ upstream(A of the ATG translation start codon is taken as +1 bp). In the present study, we have analyzed the enhancers of Otx2 expression in epiblast(EP) and anterior neuroectoderm (AN); analysis of the enhancers of Otx2 expression in forebrain and midbrain is described in an accompanying paper (Kurokawa et al.,2004). EP and AN enhancers were identified at about 90 kb 5′upstream. The EP activity was found in the inner cell mass of the blastocyst stage embryos and the entire epiblast of egg cylinder stage embryos. AN enhancer activity was detected initially at E7.0 and had ceased by E8.5. The EP enhancer included multiple required domains over 2.3 kb, and the AN enhancer was an essential component of the EP enhancer. Mutants lacking the AN enhancer were generated to demonstrate that these cis-sequences do indeed regulate Otx2 expression in EP and AN, but additional EP and AN enhancers must exist outside of the –170 kb to +120 kb range. Otx2 expression under the AN enhancer was essential to maintaining the anterior neuroectoderm once induced. Furthermore, Otx2 under the AN enhancer cooperated with Emx2 in diencephalon development.

A mouse Otx2 genomic BAC clone 582F09 (hereafter referred to as BAC#1) was isolated from a C57BL/6 BAC library (Research Genetics). It was subdivided into about 10 to 15kb fragments by Sau3AI partial digestion; the fragments were subcloned into the BamHI site of pBluescript SK(–). These clones were aligned as shown in Fig. 2A by walking. Lengths and primers used to isolate mouse genomic sequences containing the κ,λ, μ, ν, ξ, ο, Φ and χ, domains(Fig. 7A) by PCR were as follows: 3.1 kb with 5′-GCAAAACATGTTACCTGCTAAGCAAC-3′ and 5′-TGAATTCAGAGAATGCAGCGCATGC-3′; 3.3 kb with 5′-TCCTATGCAGACTTGGCTGGCCTGG-3′ and 5′-TCCCTCTCTGGCACTCAGGCTGAGG-3′; 2.8 kb with 5′-CTAGCTTTTAGCTGCAATATAGAAAGG-3′ and 5′-GGACCCTGAAGAACCGGGCAGAAAT-3′; 3.5 kb with 5′-GAGAGTAGTGCTGCATTGAACATGG-3′ and 5′-CAAGGTTGTCAAGGAGTGTTCAATG-3′; 1.6 kb with 5′-CATAAATGTGCAATGGCAAGTGTCC-3′ and 5′-CATCAGCGAGCAATCCGTAGCCAGG-3′; 1.9 kb with 5′-GAGATGATCATGCAGCTCTCGG-3′ and 5′-CTCTGTGAACCATGTAAGACTGG-3′; 0.7 kb with 5′-GAGAGAGAGGAGATTTTGCTTTCTTGAAGC-3′ and 5′-GTTCCCTTACCCTATCCACAATGCACTGTTCC-3′; and 0.7 kb with 5′-ACTATTTAACTGGCTGCCCTTTTTCGCACTGG-3′ and 5′-GAGCAGGGAGGGCATCAGATCCGCTGTACAAG-3′, respectively. The Xenopus (Silurana) tropicalis genomic sequence containing domainΔ was 1.6 kb with 5′-TGGAGAAACACAATAAAAGAGC-3′ and 5′-AACCAAAGAGTCTCCATAGC-3′.

BAC #1 reporter transgenic mice were generated as described by Helms et al.(Helms et al., 1998). The 3′ terminal of BAC #1 was located at –30 kb 5′ upstream from the Otx2 translation start site(Fig. 1A). Five kb DNA at–35 kb to –30 kb 5′ upstream was ligated to 1.8 K-lacZ, which led to the production of 5′-lacZ. BAC #1 and 5′-lacZ were separated from vector-derived sequences via digestion with NotI and SalI, respectively(Yang et al., 1997), followed by co-injection into the pronucleus of fertilized CD-1 mouse eggs(concentration of each, 1-2 ng/μl). Transgenic mice harboring the correct recombinant were identified by PCR. Primers employed for this purpose were 5′-GCCAAACACAGAATCAGCAAGGAGGAGGCCATGC-3′ in BAC #1 and 5′-CCAGATTTACATTCAGAACACAAGTCTTATTTCCCCACCC-3′ in the lacZ gene.

The 1.8 kb lacZ reporter is a modified version of the VEcis-lacZ vector of Kimura et al.(Kimura et al., 1997). The NotI-PmaCI-SacII linker was inserted into the SalI/BglII site of the 5′ polylinker in the vector. Each DNA fragment was inserted between these NotI and SacII sites. Permanent transgenic mice and transient transgenic embryos were generated and β-gal expression was determined as described previously(Kimura et al., 1997; Kimura et al., 2000).

Utilizing the PCR-based overlap extension method(Kucharczuk et al., 1999),each 15 bp block of EB165 bp AN core enhancer (#1-#11 in Fig. 4B) was replaced with the linker sequence 5′-AGACTGGATCCTGTG-3′. Putative transcription factor binding sites in each enhancer were also mutated by the transverse substitution of nucleotides using this approach.

All genomic DNAs isolated or mutagenized by PCR were sequenced to verify the absence of spurious mutations. Comparison of the mouse Otx2genomic sequences with other species was conducted with the Berkeley Genome Pipeline and Genome VISTA (Couronne et al.,2003) programs. Additionally, sequence alignment was confirmed with the BLAST program (Mayor et al.,2000). Putative transcription binding sites were predicted with the TFSEARCH program (Heinemeyer et al.,1998).

A neomycin resistance gene with Pgk1 promoter and SV40 polyadenylation signal was flanked by loxP sequences. This cassette (Neo) was replaced with the SpeI/BglII 559 bp region, which harbors the AN enhancer. Lengths of the homologous regions were 8.0 kb and 4.0 kb at the 5′ and 3′ sides of the neo cassette, respectively. Diphtheria toxin A fragment gene with MC1 promoter was used for the negative selection of homologous recombinants as described (Yagi et al., 1993b). The targeting vector was linearized with SalI and electroporated into TT2 ES cells(Yagi et al., 1993a). Homologous recombinants were identified as those that generate 4.0 kb products by PCR with 5′-ATCGCCTTCTTGACGAGTTCTTCTG-3′ in the neogene as the 5′ side primer and 5′-GATCCTCCTGCCTCTGCGTCAATAGC-3′ as the 3′ side primer; the 3′ primer is located outside of the 3′ side 4kb homologous region in Otx2 locus. The recombinants were confirmed by Southern blot analyses as described (Matsuo et al.,1995). Two mutant mouse strains were generated from independent homologous recombinant ES clones. The genotype of mutant mouse or embryo was routinely determined by PCR using tail or yolk sac specimens. Wild-type allele was detected with the sense primer of the 5′-CTGCAGATGCTTGGGCTTTCTGCAGC-3′ and the antisense primer of the 5′-GGGGTCTTCATGAGTTTCTGTTGCAC-3′. The mutant allele was identified with the sense primer of 5′-ATCGCCTTCTTGACGAGTTCTTCTG-3′ in the neo gene and the antisense primer. The deletion of the neoinsert by Cre-mediated loxp recombination was accomplished by mating Otx2^+/ΔAN mice with Lefty-Cre mice(Yamamoto et al., 2001).

Mouse embryos were fixed overnight with Bouin's fixative solution at room temperature. Specimens were subsequently dehydrated and embedded in paraplast. Serial sections (10 μm) were prepared and stained with Hematoxylin and Eosin or with 0.1% Cresyl Violet.

Whole-mount and section in situ hybridization was performed using digoxigenin probes, as described by Suda et al.(Suda et al., 2001). The probes used in this study were for the genes: brachyury/T(Herrmann, 1991), Fgf8 (Crossley and Martin,1995), Cer1 (Belo et al., 1997), Dlx1(Bulfone et al., 1993) Emx2 (Yoshida et al.,1997), Gbx2 (Bulfone et al., 1993), Lhx1/Lim1(Fujii et al., 1994), Otx2 (Matsuo et al.,1995), Pax6 (Walther and Gruss, 1991), Six3(Oliver et al., 1995) and Tcf4 (Korinek et al.,1998).

RNA isolation and reverse transcription PCR was performed according to Kimura et al. (Kimura et al.,2001). Primer sets used were as follows: Otx2,5′-TCTTATCTAAAGCAACCGCCTTACGCAGTC-3′ and 5′-GCACCCTGGATTCTGGCAAGTTGATTTTCA-3′; glyceraldehyde-3-phosphate dehydrogenase (GAPDH), 5′-TGTCATCAACGGGAAGCCCA-3′ and 5′-TTGTCATGGATGACCTTGGC-3′.

Animals were housed in environmentally controlled rooms in accordance with the Institute of Physical and Chemical Research (RIKEN) guidelines for animal experiments.

We had previously searched the mouse genomic region from –50 kb 5′ upstream to +10 kb 3′ downstream of the Otx2translation start site for sequences involved in regulation of Otx2expression (Kimura et al.,1997; Kimura et al.,2000) (C.K.-Y., I.M. and S.A., unpublished). The major transcription start site of Otx2 expression exists at –207 bp 5′ upstream of the translational start site(Guazzi et al., 1998; Courtois et al., 2003). The–1.8 kb 5′ region adjacent to the translational start site is capable of directing the expression in VE, AME and CM(Kimura et al., 1997; Kimura et al., 2000); the construct in which the 1.8 kb genome was combined with the lacZ gene was designated as 1.8k-lacZ. However, no sequences were detected which directed Otx2 expression in epiblast, anterior neuroectoderm or forebrain/midbrain. In the present study, we generated transgenic mice harboring a genomic DNA spanning –170 kb to –30 kb 5′upstream (BAC #1) with 1.8k-lacZ(Fig. 1A) in order to identify these sequences. The 5′ BAC#1/lacZ transgene was generated by homologous recombination in mouse zygotes(Helms and Johnson, 1998) by co-injecting BAC#1 and a lacZ reporter; the reporter had the 5kb 3′ end of BAC#1 ligated to the 1.8k-lacZ.

The 5′BAC#1/lacZ transgenic mouse line capturedβ-gal expression in E6.5 epiblast, E7.75 anterior neuroectoderm and E9.5 forebrain and anterior midbrain (Fig. 1B, parts a-c). β-Gal expression was not detected in posterior midbrain (Fig. 1B,part c). Subsequently, the Otx2 genomic region encompassing–170 kb to –30 kb was systematically dissected into 15 fragments with partial overlaps at both ends to determine the locations of regulatory elements (Fig. 2A). All enhancer analyses described below were conducted using the mouse Otx21.8 kb promoter. Expression by this promoter(Fig. 1B, part d) is indicated by arrows in each figure. For several DNA fragments, enhancer activities were also assayed using Fugu Otx2 promoter(Kimura et al., 1997) or hsp68 promoter. For example, an EcoRV/BglII 2.3kb fragment capable of driving lacZ expression in epiblast with the 1.8 kb promoter (Fig. 3B, parts b-d) exhibited the activity, though weak, using the hsp68 promoter(Fig. 3B, parts e,f). An enhancer in the #13 fragment (Fig. 2A) that had the transcriptional activity in forebrain and midbrain with the 1.8 kb promoter showed a similar activity with the Fugu promoter. By contrast, a SpeI 6.2 kb fragment(Fig. 3A) that exhibited activity in the anterior neuroectoderm did not show the same activity with the Fugu promoter, and exhibited low activity in the ectoderm and ectopic activity in extra-embryonic ectoderm with the hsp68 promoter (data not shown). Moreover, a EcoRV/BglII 2.3 kb fragment(Fig. 3A) exhibited similar activities in the anterior neuroectoderm with the hsp68 promoter and mouse 1.8 kb promoter (Fig. 3B,parts g,h). Thus, the enhancer activity of several DNA fragments described below is promoter dependent. The details of the promoter dependence, however,remain a topic for future studies. Note that, in the present paper, the term`enhancer' is used in the broad sense of `cis-regulatory sequence'.

Transgenic lines were established for each of the 15 fragments as indicated in the parenthetical statement in Fig. 2A. Analyses of these lines mapped the enhancer of Otx2expression in E6.5 epiblast and E7.75 anterior neuroectoderm to fragment #12,which was located about –80 kb to –95 kb 5′ upstream. The enhancer of Otx2 expression in E9.5 forebrain and midbrain was mapped to fragment #13, which was located about –70 kb to –80 kb 5′upstream. In this mapping, heterozygous Otx2^+/lacZ embryos(Kimura et al., 2000) were stained in parallel to identify β-gal expression from the endogenous Otx2 locus (Fig. 2B,parts g-k). When necessary, whole-mount RNA in situ hybridization analysis was performed in transgenic embryos to compare endogenous Otx2 and transgenic lacZ expression. These analyses unequivocally indicated that fragments #12 and #13 contain sequences necessary to direct Otx2expression in ectoderm at each developmental stage. This study focuses on the enhancer activity in the #12 fragment; an analysis of the activity of the #13 fragment is reported (Kurokawa et al.,2004). In addition to #12 and #13 fragments, the #4 fragment droveβ-gal expression in a region of the ventral diencephalon and dorsal mesencephalon, the significance of which will be addressed in future studies(Fig. 2B, part l). Some fragments exhibited β-gal expression in structures where Otx2 is not expressed endogenously; no further analysis was made of these activities.

The activity of the #12 fragment was initially noted in entire epiblast; at E6.5 it was absent in the posterior region(Fig. 2B, parts a,b). At E7.75,β-gal expression was restricted to the anterior neuroectoderm, which corresponds to the presumptive forebrain and midbrain(Fig. 2B, parts c,d); this activity was lost by E8.5 (Fig. 2B, part e). After E9.0, the fragment exhibited activity in the most dorsomedial aspect of the telencephalon and in the eyes(Fig. 2B, part f). From this activity profile, the enhancer in the #12 fragment was designated as the EP/AN(epiblast/anterior neuroectoderm) enhancer.

The expression of β-gal in the epiblast at the Otx2 locus is noteworthy. Endogenous Otx2 is expressed in epiblast. However,β-gal expression is not observed in the epiblast of Otx2^+/lacZ embryos(Fig. 2B, part g)(Boyl et al., 2001; Courtois et al., 2003). By contrast, β-gal expression was evident in epiblast with the BAC#1/lacZ transgene (Fig. 1B, part a) and with the lacZ transgene under the #12 fragment (Fig. 2B, parts a,b),respectively. The #12 fragment also drove β-gal expression in epiblast under the Otx2^+/– background (data not shown); the discrepancy is not due to the Otx2 dose effect. Translational control has been suggested in the expression of transgenes knocked-in into Otx2 locus (Boyl et al.,2001). However, in both the lacZ knock-in mutation into the Otx2 locus and the random lacZ transgenesis in the enhancer analysis, the lacZ gene was fused to the Otx25′ untranslated region and SV40 3′UTR in exactly the same manner. The cause of these variations of β-gal expression in epiblast is unclear.

The 15 kb #12 fragment was subdivided by SpeI digestion(Fig. 3A) in order to define the EP/AN enhancer. The enhancer activity of each subfragment was determined by transient transgenic assay at E6.5 and E7.75. However, none of these SpeI fragments drove β-gal expression in epiblast at E6.5. Subsequently, the #12 fragment was re-subdivided into five fragments so that each SpeI site was preserved (Fig. 3A). In this subdivision, β-gal expression in epiblast was driven by the EcoRV/BglII 2.3 kb fragment(Fig. 3A), which was confirmed by generation of permanent transgenic lines. Activity of the EcoRV/BglII 2.3 kb fragment was initially found in the inner cell mass of blastocysts at E3.5 (Fig. 3B, part a) and in the entire epiblast at E5.5(Fig. 3B, part b), and gradually declined in the posterior region with the progress of gastrulation in a manner similar to endogenous Otx2 expression(Fig. 3B, parts c-f). At the headfold stage the fragment exhibited the activity in anterior neuroectoderm(Fig. 3B, parts g,h).

Unexpectedly, however, when the EcoRV/BglII 2.3 kb fragment was subdivided further, fragments that lacked any of the following sequences lost their activity: (1) EcoRV/HindIII 0.95 kb,(2) ScaI/NspI 0.6 kb, (3) NspI/EcoT22I 0.57 kb or (4) EcoT22I/BglII 165 bp(Fig. 3A). Thus, all of these domains were essential for enhancer activity in epiblast. Regions homologous to the EcoRV/BglII 2.3 kb fragment exist at similar positions relative to the Otx2 coding region in the rat, human and Xenopus (Silurana) tropicalis (hereafter, with regard to the genome information, `Xenopus' refers to this species)genomes (see Fig. 7). Regions displaying greater than 80% sequence identity over more than 100 bp are present throughout the EcoRV/BglII 2.3 kb fragment(Fig. 3A).

In contrast to the epiblast enhancer, the regulatory sequences of Otx2 expression in anterior neuroectoderm were present in the SpeI 6.2 kb fragment (Fig. 3A). This fragment was successively dissected as shown in Fig. 4A, and the transient transgenic assay at E7.75 revealed the existence of activity exclusively in the SpeI/BglII 559 bp. With this fragment β-gal expression in the anterior neuroectoderm was detected initially at E7.0 and ceased by E8.5. The caudal boundary of β-gal expression was not distinct near the preotic sulcus (Fig. 4C, part a; data not shown).

The SpeI/BglII 559 bp fragment was further dissected, and finally the EcoT22I/BglII 165 bp fragment was found to retain the enhancer activity; this was confirmed by generation of permanent transgenic lines (Fig. 4C, part a). A region exhibiting 92% identity to this 165 bp fragment is found about 5′ 97 kb upstream in the human Otx2 locus(Fig. 4B; see Fig. 7A). The 5′ part of the 165 bp fragment includes the sequences TGGCGACTGAC and TGGGCGCTGGC, which correspond to each other with the exception of two bases. Furthermore, these sequences are conserved in the human counterpart. Five tandem repeats of these sequences, however, did not direct β-gal expression in anterior neuroectoderm (Fig. 4A). In the 165 bp fragment, potential OCT/HOX (Herr and Cleary, 1995) and SOX (van Beest et al., 2000) binding sites exist; however, AN activity was unaffected by mutations in these sites.

A linker-scanner approach was then employed to define the core elements. Each 15 bp sequence of the EcoT22I/BglII 165 bp fragment was replaced with the BamHI linker (Fig. 4B) (Kucharczuk et al.,1999). Replacement of any of the six 5′ side 15 bp blocks(#6-#11) affected β-gal expression in the anterior neuroectoderm at E7.75. Such replacements in three of these blocks (mt#6, mt#7 and mt#10)essentially abolished the activity (Fig. 4C, part b), whereas mutations in #8, #9 and #11 demonstrated faint β-gal expression (Fig. 4C, part c) at the frequencies indicated in the parentheses of Fig. 4B (asterisks). In addition, when 75 bp sequences corresponding to #11-7 were duplicated in tandem and combined with 1.8 k-lacZ, the transgene exhibited significant β-gal expression in E7.75 anterior neuroectoderm(Fig. 4C, part d). A region homologous to EcoT22I/BglII 165 bp is present about 5′ 44 kb upstream in the Xenopus Otx2 locus(Fig. 7, Fig. 4B), displaying sequence identity of 77%. Identity is 87% in the #11-#6 90 bp sequences; however, no homologous region is present in the zebrafish or pufferfish genomes. A 1.6 kb Xenopus DNA region covering this domain was then isolated by PCR, and its enhancer activity was tested using the mouse 1.8 kb promoter. The 1.6 kb region clearly directed β-gal expression in mouse anterior neuroectoderm(Fig. 4C, part e)

The analysis described above indicated that the SpeI/BglII 559 bp region was essential for expression in epiblast, though not, by itself, sufficient. This fragment is located ∼90 kb away from the coding region, which raised a question regarding its involvement in the regulation of Otx2 expression in vivo. In order to examine this point, mutant mice (Otx2^ΔAN/ΔAN)were generated in which the SpeI/BglII 559 bp was replaced with a cassette encoding a neomycin-resistant gene(Fig. 5A).

The epiblast develops in Otx2-null mutant(Otx2^–/–) embryos(Kimura et al., 2000), and it also developed in Otx2^ΔAN/ΔAN embryos. In this epiblast, the Otx2 expression was diminished but not abolished(Fig. 5B, parts a,b). Quantitative RT-PCR confirmed a reduction in Otx2 expression(Fig. 5B, part c), indicating that the SpeI/BglII 559 bp region functions as an epiblast enhancer, but also that another enhancer of Otx2 expression in epiblast must be present.

In mammals, the anteroposterior axis is first generated along the distoproximal axis. Prior to primitive streak formation, distal visceral endoderm cells move to the future anterior and proximal epiblast cells to the future posterior. This axis rotation does not occur in Otx2^–/– mutants(Kimura et al., 2000). However, axis rotation proceeded normally in Otx2^ΔAN/ΔAN mutants as well as in Otx2^ΔAN/– embryos, in which one Otx2allele lacks the SpeI/BglII 500 bp and the other allele is null, as indicated by cerberus-like expression in the anterior visceral endoderm (Fig. 5C,part a) and T expression in the primitive streak(Fig. 5C, part b) at E6.5. This observation is consistent with our previous conclusion that Otx2-positive distal visceral endoderm cells, but not Otx2-positive epiblast, are responsible for the axis rotation defect in Otx2^–/– mutants(Kimura et al., 2000).

The SpeI/BglII 559 bp was essential and sufficient for driving Otx2 expression in anterior neuroectoderm. In Otx2^ΔAN/ΔAN mutants, however, no defects were found in the forebrain or midbrain. Otx2 expression was reduced but not eliminated in E7.5 Otx2^ΔAN/ΔAN mutants(data not shown). Subsequently, Otx2 levels were reduced further via the generation of Otx2^ΔAN/– mutants. At E12.5 the Otx2^ΔAN/– mutant phenotype was variable. In milder cases forebrain was present though greatly deformed; in severe cases both forebrain and midbrain were lost (Fig. 6A, part b). At E7.5 Six3-positive anterior neuroectoderm was induced in all Otx2^ΔAN/– mutants examined(Fig. 6B, part b). At E8.5,however, Six3-positive region was reduced(Fig. 6B, part d), and concomitantly Fgf8- or Gbx2-negative anterior neuroectoderm was reduced but still present at E8.5 (Fig. 6B, parts f,h). The phenotype became variable at E9.5, and in severe mutants Gbx2-negative forebrain/midbrain and Emx2-positive forebrain were entirely missing(Fig. 6B, parts j, l). Thus,the SpeI/BglII 559 bp located 90 kb 5′ upstream clearly regulates Otx2 expression in anterior neuroectoderm; however,another enhancer must also exist.

In our previous analysis of Otx2^+/–Emx2^–/– mutants, we reported that Otx2 cooperates with Emx2 in diencephalon development (Suda et al.,2001). Emx2 expression occurs at the three-somite stage(E8.0) when the AN enhancer is active. To confirm that Emx2cooperates with Otx2 under the AN enhancer, Otx2^ΔAN/ΔANEmx2^–/–mutants were generated. In these mutants, diencephalon was lost(Fig. 6A, part c); at E11.5 Pax6- and Dlx1-positive ventral thalamus and Tcf4-and Gbx2-positive dorsal thalamus were absent(Fig. 6C, parts a-h), while Lhx1- and Tcf4-positive commissural region of the pretectum developed (Fig. 6C, parts g-j). This phenotype was very similar to the Otx2^+/–Emx2^–/– mutant phenotype.

Enhancer mutants indicated the presence of another enhancer of Otx2 expression in epiblast and anterior neuroectoderm. Consequently,a search was conducted for the second enhancer in the 3′ region of the Otx2 gene. BAC#2 transgenic mice, which harbor a genomic DNA region spanning –50 kb to +120 kb 3′ downstream, were generated(Fig. 7)(Kurokawa et al., 2004). However, this region displayed no activity in epiblast or anterior neuroectoderm.

Mouse Otx2 locates at chromosome14 19 cM; the nearest genes are encoded ∼350 kb upstream on the 5′ side and ∼150 kb downstream on the 3′ side (Fig. 7). More than 100 highly conserved domains (sequence identity greater than 80%over more than 100 bp) exist in this 500 kb span among mouse, rat and human. Of these domains, 22 are also conserved in the Xenopus genome. Fig. 7 depicts the locations of these domains in the mouse, human and Xenopus Otx2 loci. The organization is identical in mouse and human. The contiguous mapping is partial in the Xenopus Otx2 locus; however, domains α-ι,ρ and σ locate in identical order in a more compact fashion.

The EP/AN enhancer region identified above was one of the 22 domains conserved, Δ. We next sought a second EP/AN enhancer among domainsκ-ο, Φ and χ, which exist outside of the regions analyzed with BAC#1 and #2. Depending on the sizes of the conserved regions, the lengths of genome DNAs isolated by PCR were 3.1, 3.3, 2.8, 3.5, 1.6, 1.9, 0.7 and 0.7 kb for κ, λ, μ, ν, ξ, ο, Φ and χdomains, respectively. Subsequently, sequences were confirmed and enhancer activity was determined with the 1.8 kb promoter by transient transgenic assay. None of these domains, however, exhibited enhancer activity in E6.5 epiblast or E7.75 anterior neuroectoderm.

We previously identified enhancers of Otx2 expression in distal/anterior visceral endoderm (DVE/AVE), anterior mesendoderm (AME) and cephalic mesenchyme (CM) (Kimura et al.,1997; Kimura et al.,2000). In the present study, we have identified enhancers of Otx2 expression in epiblast (EP) and anterior neuroectoderm (AN);Kurokawa et al. (Kurokawa et al.,2004) report enhancers of Otx2 expression in forebrain/midbrain (FM). DVE/AVE, AME and CM enhancers located near the Otx2-coding region; by contrast, all ectodermal enhancers occurred at great distances from the coding region.

Otx2 expression is differentially regulated in each germ layer. The regulation of these expressions by different enhancers could be reasonably expected; the existence of distinct VE, AME, CM and ectoderm enhancers was anticipated. In mouse, Otx2 expression in epiblast and anterior neuroectoderm is continuous. However, Otx2 expression in chick and in Xenopus is consistent with discrete EP and ANE enhancers. In avian,following the disappearance of Otx2 expression in epiblast,expression occurs transiently in Hensen's node and in the anterior primitive streak exclusively and appears in the anterior neuroectoderm subsequently(Bally-Cuif et al., 1995; Kablar et al., 1996). AN enhancer activity begins around E7.0 and ceases by E8.5. Of note is that the EP enhancer also directed the expression in inner cell mass of blastocysts.

Generally, the identification of enhancers is conducted solely by random transgenesis. This study unequivocally demonstrated the necessity of mutating the enhancer in vivo to determine whether and how the enhancer is responsible for endogenous gene expression. The EP/AN enhancer mutants indicated that another enhancer exists for Otx2 expression in epiblast and anterior neuroectoderm. However, the second enhancer was not present within the 3′ 120 kb region covered by BAC #2. The overall sequences in EP and AN enhancer region are well conserved in mouse, human and Xenopus. Then the second EP/AN enhancers were searched for among domains conserved in mouse,human and Xenopus, outside of the 290 kb region covered by BAC#1 and BAC#2. However, no EP/AN activity was detected in any of these domains. The second enhancer might exist much farther away over the adjacent genes. At the same time, the possibility remains that the second EP/AN enhancer exists uniquely in rodent outside of the 290 kb genomic region, as the second enhancer of Otx2 expression in forebrain and midbrain, FM2, is not conserved in human or Xenopus, and is most probably unique to the rodent (Kurokawa et al.,2004).

The EP enhancer appears to consist of multiple domains, and the AN enhancer is a component of the EP enhancer, the significance of which remains to be addressed in future studies. EP and AN enhancer regions are conserved in human and Xenopus, a Xotx2 1.6 kb region covering the AN enhancer region exhibited the enhancer activity in mouse anterior neuroectoderm. The EP/AN region is, however, not conserved in zebrafish or pufferfish. In zebrafish blastoderm Otx1, but not Otx2, is expressed. Fish blastoderm may be a product independent of the amniote epiblast. However, the lack of conservation of the AN enhancer region in fish genome was unexpected. Otx2 alone in tetrapod and all Otx genes in zebrafish are expressed in the anterior neuroectoderm. It is possible that the ancient AN enhancer widely diverged between the lineages to the extant teleost and tetrapod to cope with the divergence in anterior neuroectoderm development, and so the molecular machinery involved in its development might be extremely varied between teleost and tetrapod (Kurokawa et al., 2004). Discussion of the phylogenetic significance of enhancer organization obviously requires the identification of the second EP/AN enhancer. In addition, a comprehensive enhancer analysis must be performed in fish Otx genes (Kimura et al., 1997; Kimura-Yoshida et al., 2004).

The generation of enhancer mutants to allow the dissection of Otx2functions in each site at each step of head development was a primary objective of enhancer analysis in the present investigation. The presence of two enhancers of Otx2 expression in anterior neuroectoderm complicated the analysis. Nevertheless, Otx2^ΔAN/– mutants clearly demonstrated that Otx2 expression under the AN enhancer functioned to maintain the anterior neuroectoderm once induced. This was originally suggested by analyses of Otx1 knock-in mutants into Otx2 locus(Otx2^Otx1/Otx1) and a hypomorphic Otx2 mutant(Otx2^{frt-neo/frt-neo})(Suda et al., 1999; Li and Joyner, 2001; Martinez-Barbera et al., 2001; Tian et al., 2002). However,in these mutants the role of Otx2 expression in anterior mesendoderm in the loss of the anterior neuroectoderm could not be ruled out.

Studies of Otx2^+/–Emx2^–/–mutants have previously suggested that Otx2 cooperates with Emx2 in the development of the diencephalic region(Suda et al., 2001). However,the onset of the defects was not clear. Otx2^ΔAN/ΔANEmx2^–/–phenotypes were quite similar to Otx2^+/–Emx2^–/– phenotypes; at E12.5 ventral thalamus, dorsal thalamus and anterior part of the pretectum were lost, but the commissural region of the pretectum had developed. No marker analyses were made, but histologically archipallium was not apparent,while neopallium and ganglionic eminences developed. Mesencephalon was largely normal, while cerebellum was enlarged in the double mutants. Emx2expression takes place in the laterocaudal forebrain primordium around the three-somite stage (E8.0) when the AN enhancer is active. Furthermore,β-gal expression under the AN enhancer was not detected at E8.5;β-gal is stable and usually persists longer than endogenous gene products. Thus, we propose that the diencephalic region may be induced around the three-somite stage by cooperation between Emx2 and Otx2under the AN enhancer. With loss of this cooperation, the structures that derive from the Emx2-positive region at the three-somite stage may be lost.

We thank Ms S. Nagayoshi, T. Ohmura and Mr. N. Takeda for their technical assistance. We are indebted to Drs E. M. De Robertis, M. A. Frohman, N. Heintz, B. G. Herrmann, P. Gruss and G. R. Martin, who generously provided us a number of plasmids, and to Dr H. Hamada, who kindly provided us Lefty-Cre mice. We are also grateful to the Laboratory for Animal Resource and Genetic Engineering for the housing of mice. This work was supported by a Grant-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Culture, Sports, Science and Technology of Japan.